{ "args": { "variant_name": "pilot_primitive_router", "seed": 83, "device": "auto", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "primitive_router", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "init_logit_scale": 8.0, "temperature": 1.0, "train_min_len": 1, "train_max_len": 6, "eval_lengths": "3,6,9,12", "eval_k": "0,1,2,3,6,9,12", "eval_query_types": "all", "train_steps": 500, "batch_size": 256, "eval_batch_size": 256, "eval_examples": 512, "lr": 0.001, "weight_decay": 0.0001, "belief_loss_weight": 1.0, "query_loss_weight": 0.0, "slot_entropy_weight": 0.0, "max_grad_norm": 1.0, "log_every": 100, "output_dir": "experiments/cyclic_transition_ladder/runs/pilot_primitive_router", "checkpoint_dir": "large_artifacts/cyclic_transition_ladder/checkpoints/pilot_primitive_router" }, "train": [ { "step": 1, "loss": 3.9729790687561035, "grad_norm": 0.23458124697208405, "elapsed_seconds": 0.6569972038269043, "mean_slot_purity": 0.13658583164215088, "mean_weight_entropy": 1.9862065315246582, "mean_route_entropy": 2.30029559135437, "mean_route_accuracy": 0.12890625, "belief_loss": 3.9729790687561035, "decoder_belief_target_mass": 0.08928534388542175, "decoder_belief_target_nll": 4.52941370010376, "decoder_belief_top1_on_support": 0.16796875, "mean_decoder_belief_support_size": 5.5234375, "decoder_query_target_mass": 0.45648136734962463, "decoder_query_target_nll": 2.3482608795166016, "decoder_query_top1_on_support": 0.515625, "mean_decoder_query_support_size": 4.734375 }, { "step": 100, "loss": 1.7639589309692383, "grad_norm": 0.001859114971011877, "elapsed_seconds": 26.062679290771484, "mean_slot_purity": 0.75770503282547, "mean_weight_entropy": 1.3663818836212158, "mean_route_entropy": 0.0010096216574311256, "mean_route_accuracy": 1.0, "belief_loss": 1.7639589309692383, "decoder_belief_target_mass": 0.9996597766876221, "decoder_belief_target_nll": 1.3664252758026123, "decoder_belief_top1_on_support": 1.0, "mean_decoder_belief_support_size": 5.390625, "decoder_query_target_mass": 0.9998152256011963, "decoder_query_target_nll": 1.1530537605285645, "decoder_query_top1_on_support": 1.0, "mean_decoder_query_support_size": 4.7265625 }, { "step": 200, "loss": 1.6925442218780518, "grad_norm": 0.0008642351604066789, "elapsed_seconds": 51.31000995635986, "mean_slot_purity": 0.7278753519058228, "mean_weight_entropy": 1.4112749099731445, "mean_route_entropy": 0.0006891153170727193, "mean_route_accuracy": 1.0, "belief_loss": 1.6925442218780518, "decoder_belief_target_mass": 0.9998201727867126, "decoder_belief_target_nll": 1.4113112688064575, "decoder_belief_top1_on_support": 1.0, "mean_decoder_belief_support_size": 5.58984375, "decoder_query_target_mass": 0.9999110102653503, "decoder_query_target_nll": 1.2159379720687866, "decoder_query_top1_on_support": 1.0, "mean_decoder_query_support_size": 5.0234375 }, { "step": 300, "loss": 1.7841219902038574, "grad_norm": 0.0005664927302859724, "elapsed_seconds": 76.80283617973328, "mean_slot_purity": 0.7327370643615723, "mean_weight_entropy": 1.4062087535858154, "mean_route_entropy": 0.00048785872058942914, "mean_route_accuracy": 1.0, "belief_loss": 1.7841219902038574, "decoder_belief_target_mass": 0.9998583197593689, "decoder_belief_target_nll": 1.406141996383667, "decoder_belief_top1_on_support": 1.0, "mean_decoder_belief_support_size": 5.359375, "decoder_query_target_mass": 0.9999229311943054, "decoder_query_target_nll": 1.147935152053833, "decoder_query_top1_on_support": 1.0, "mean_decoder_query_support_size": 4.5234375 }, { "step": 400, "loss": 1.7864991426467896, "grad_norm": 0.0004515537293627858, "elapsed_seconds": 102.17790102958679, "mean_slot_purity": 0.7523543834686279, "mean_weight_entropy": 1.42649507522583, "mean_route_entropy": 0.00039953674422577024, "mean_route_accuracy": 1.0, "belief_loss": 1.7864991426467896, "decoder_belief_target_mass": 0.9998914003372192, "decoder_belief_target_nll": 1.4264659881591797, "decoder_belief_top1_on_support": 1.0, "mean_decoder_belief_support_size": 5.68359375, "decoder_query_target_mass": 0.9999508857727051, "decoder_query_target_nll": 1.2219929695129395, "decoder_query_top1_on_support": 1.0, "mean_decoder_query_support_size": 4.984375 }, { "step": 500, "loss": 1.7801659107208252, "grad_norm": 0.0003114561550319195, "elapsed_seconds": 127.67007613182068, "mean_slot_purity": 0.7528151273727417, "mean_weight_entropy": 1.4524359703063965, "mean_route_entropy": 0.0003697958018165082, "mean_route_accuracy": 1.0, "belief_loss": 1.7801659107208252, "decoder_belief_target_mass": 0.9999240040779114, "decoder_belief_target_nll": 1.452406406402588, "decoder_belief_top1_on_support": 1.0, "mean_decoder_belief_support_size": 5.828125, "decoder_query_target_mass": 0.9999603033065796, "decoder_query_target_nll": 1.2140852212905884, "decoder_query_top1_on_support": 1.0, "mean_decoder_query_support_size": 5.0703125 } ], "eval": [ { "step": 500, "rows": [ { "model": "cyclic_transition_ladder", "variant": "pilot_primitive_router", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "primitive_router", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 500, "length": 3, "query_type": "A", "k": 0, "n": 512, "decoder_query_target_mass": 0.46342329680919647, "decoder_query_target_nll": 2.397895336151123, "decoder_query_top1_on_support": 0.482421875, "mean_decoder_query_support_size": 5.09765625, "decoder_belief_target_mass": 0.06605108827352524, "decoder_belief_target_nll": 15.732054233551025, "decoder_belief_top1_on_support": 0.1015625, "mean_decoder_belief_support_size": 5.849609375, "mean_slot_purity": 0.999997615814209, "mean_slot_entropy": 3.839540295302868e-05, "mean_weight_entropy": 2.397895336151123, "mean_route_entropy": -1.0, "mean_route_accuracy": -1.0, "mean_target_support_size": 5.849609375 }, { "model": "cyclic_transition_ladder", "variant": "pilot_primitive_router", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "primitive_router", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 500, "length": 3, "query_type": "A", "k": 1, "n": 512, "decoder_query_target_mass": 0.5811124444007874, "decoder_query_target_nll": 4.298899412155151, "decoder_query_top1_on_support": 0.587890625, "mean_decoder_query_support_size": 5.09765625, "decoder_belief_target_mass": 0.15382160246372223, "decoder_belief_target_nll": 13.567691802978516, "decoder_belief_top1_on_support": 0.203125, "mean_decoder_belief_support_size": 5.849609375, "mean_slot_purity": 0.8649760484695435, "mean_slot_entropy": 0.21148967742919922, "mean_weight_entropy": 1.9719488620758057, "mean_route_entropy": 0.0003308418672531843, "mean_route_accuracy": 1.0, "mean_target_support_size": 5.849609375 }, { "model": "cyclic_transition_ladder", "variant": "pilot_primitive_router", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "primitive_router", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 500, "length": 3, "query_type": "A", "k": 2, "n": 512, "decoder_query_target_mass": 0.7185370922088623, "decoder_query_target_nll": 3.789208769798279, "decoder_query_top1_on_support": 0.71875, "mean_decoder_query_support_size": 5.09765625, "decoder_belief_target_mass": 0.27200931310653687, "decoder_belief_target_nll": 10.28016471862793, "decoder_belief_top1_on_support": 0.3125, "mean_decoder_belief_support_size": 5.849609375, "mean_slot_purity": 0.79668128490448, "mean_slot_entropy": 0.3689309060573578, "mean_weight_entropy": 1.708378553390503, "mean_route_entropy": 0.0003564246144378558, "mean_route_accuracy": 1.0, "mean_target_support_size": 5.849609375 }, { "model": "cyclic_transition_ladder", "variant": "pilot_primitive_router", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "primitive_router", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 500, "length": 3, "query_type": "A", "k": 3, "n": 512, "decoder_query_target_mass": 0.9999804794788361, "decoder_query_target_nll": 1.2583229541778564, "decoder_query_top1_on_support": 1.0, "mean_decoder_query_support_size": 5.09765625, "decoder_belief_target_mass": 0.9999346137046814, "decoder_belief_target_nll": 1.4816256165504456, "decoder_belief_top1_on_support": 1.0, "mean_decoder_belief_support_size": 5.849609375, "mean_slot_purity": 0.7511433064937592, "mean_slot_entropy": 0.5015242695808411, "mean_weight_entropy": 1.4816245436668396, "mean_route_entropy": 0.00033705588430166245, "mean_route_accuracy": 1.0, "mean_target_support_size": 5.849609375 }, { "model": "cyclic_transition_ladder", "variant": "pilot_primitive_router", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "primitive_router", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 500, "length": 3, "query_type": "A", "k": 6, "n": 512, "decoder_query_target_mass": 0.9999804794788361, "decoder_query_target_nll": 1.2583229541778564, "decoder_query_top1_on_support": 1.0, "mean_decoder_query_support_size": 5.09765625, "decoder_belief_target_mass": 0.9999346137046814, "decoder_belief_target_nll": 1.4816256165504456, "decoder_belief_top1_on_support": 1.0, "mean_decoder_belief_support_size": 5.849609375, "mean_slot_purity": 0.7511433064937592, "mean_slot_entropy": 0.5015242695808411, "mean_weight_entropy": 1.4816245436668396, "mean_route_entropy": -1.0, "mean_route_accuracy": -1.0, "mean_target_support_size": 5.849609375 }, { "model": "cyclic_transition_ladder", "variant": "pilot_primitive_router", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "primitive_router", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 500, "length": 3, "query_type": "A", "k": 9, "n": 512, "decoder_query_target_mass": 0.9999804794788361, "decoder_query_target_nll": 1.2583229541778564, "decoder_query_top1_on_support": 1.0, "mean_decoder_query_support_size": 5.09765625, "decoder_belief_target_mass": 0.9999346137046814, "decoder_belief_target_nll": 1.4816256165504456, "decoder_belief_top1_on_support": 1.0, "mean_decoder_belief_support_size": 5.849609375, "mean_slot_purity": 0.7511433064937592, "mean_slot_entropy": 0.5015242695808411, "mean_weight_entropy": 1.4816245436668396, "mean_route_entropy": -1.0, "mean_route_accuracy": -1.0, "mean_target_support_size": 5.849609375 }, { "model": "cyclic_transition_ladder", "variant": "pilot_primitive_router", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "primitive_router", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 500, "length": 3, "query_type": "A", "k": 12, "n": 512, "decoder_query_target_mass": 0.9999804794788361, "decoder_query_target_nll": 1.2583229541778564, "decoder_query_top1_on_support": 1.0, "mean_decoder_query_support_size": 5.09765625, "decoder_belief_target_mass": 0.9999346137046814, "decoder_belief_target_nll": 1.4816256165504456, "decoder_belief_top1_on_support": 1.0, "mean_decoder_belief_support_size": 5.849609375, "mean_slot_purity": 0.7511433064937592, "mean_slot_entropy": 0.5015242695808411, "mean_weight_entropy": 1.4816245436668396, "mean_route_entropy": -1.0, "mean_route_accuracy": -1.0, "mean_target_support_size": 5.849609375 }, { "model": "cyclic_transition_ladder", "variant": "pilot_primitive_router", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "primitive_router", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 500, "length": 3, "query_type": "B", "k": 0, "n": 512, "decoder_query_target_mass": 0.455078125, "decoder_query_target_nll": 2.397895336151123, "decoder_query_top1_on_support": 0.439453125, "mean_decoder_query_support_size": 5.005859375, "decoder_belief_target_mass": 0.07120022177696228, "decoder_belief_target_nll": 15.447587013244629, "decoder_belief_top1_on_support": 0.119140625, "mean_decoder_belief_support_size": 5.912109375, "mean_slot_purity": 0.999997615814209, "mean_slot_entropy": 3.839540295302868e-05, "mean_weight_entropy": 2.397895336151123, "mean_route_entropy": -1.0, "mean_route_accuracy": -1.0, "mean_target_support_size": 5.912109375 }, { "model": "cyclic_transition_ladder", "variant": "pilot_primitive_router", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "primitive_router", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 500, "length": 3, "query_type": "B", "k": 1, "n": 512, "decoder_query_target_mass": 0.5561955273151398, "decoder_query_target_nll": 4.046899199485779, "decoder_query_top1_on_support": 0.52734375, "mean_decoder_query_support_size": 5.005859375, "decoder_belief_target_mass": 0.14056428521871567, "decoder_belief_target_nll": 13.530315399169922, "decoder_belief_top1_on_support": 0.17578125, "mean_decoder_belief_support_size": 5.912109375, "mean_slot_purity": 0.8856020867824554, "mean_slot_entropy": 0.18314127624034882, "mean_weight_entropy": 2.059309244155884, "mean_route_entropy": 0.0003417999396333471, "mean_route_accuracy": 1.0, "mean_target_support_size": 5.912109375 }, { "model": "cyclic_transition_ladder", "variant": "pilot_primitive_router", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "primitive_router", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 500, "length": 3, "query_type": "B", "k": 2, "n": 512, "decoder_query_target_mass": 0.7325931787490845, "decoder_query_target_nll": 3.78212833404541, "decoder_query_top1_on_support": 0.734375, "mean_decoder_query_support_size": 5.005859375, "decoder_belief_target_mass": 0.2639910727739334, "decoder_belief_target_nll": 10.552316188812256, "decoder_belief_top1_on_support": 0.310546875, "mean_decoder_belief_support_size": 5.912109375, "mean_slot_purity": 0.8033351898193359, "mean_slot_entropy": 0.373907670378685, "mean_weight_entropy": 1.703046202659607, "mean_route_entropy": 0.0003316258662380278, "mean_route_accuracy": 1.0, "mean_target_support_size": 5.912109375 }, { "model": "cyclic_transition_ladder", "variant": "pilot_primitive_router", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "primitive_router", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 500, "length": 3, "query_type": "B", "k": 3, "n": 512, "decoder_query_target_mass": 0.9999813437461853, "decoder_query_target_nll": 1.2284263372421265, "decoder_query_top1_on_support": 1.0, "mean_decoder_query_support_size": 5.005859375, "decoder_belief_target_mass": 0.9999337196350098, "decoder_belief_target_nll": 1.5004963874816895, "decoder_belief_top1_on_support": 1.0, "mean_decoder_belief_support_size": 5.912109375, "mean_slot_purity": 0.7600073218345642, "mean_slot_entropy": 0.49823057651519775, "mean_weight_entropy": 1.5004939436912537, "mean_route_entropy": 0.0003426629991736263, "mean_route_accuracy": 1.0, "mean_target_support_size": 5.912109375 }, { "model": "cyclic_transition_ladder", "variant": "pilot_primitive_router", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "primitive_router", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 500, "length": 3, "query_type": "B", "k": 6, "n": 512, "decoder_query_target_mass": 0.9999813437461853, "decoder_query_target_nll": 1.2284263372421265, "decoder_query_top1_on_support": 1.0, "mean_decoder_query_support_size": 5.005859375, "decoder_belief_target_mass": 0.9999337196350098, "decoder_belief_target_nll": 1.5004963874816895, "decoder_belief_top1_on_support": 1.0, "mean_decoder_belief_support_size": 5.912109375, "mean_slot_purity": 0.7600073218345642, "mean_slot_entropy": 0.49823057651519775, "mean_weight_entropy": 1.5004939436912537, "mean_route_entropy": -1.0, "mean_route_accuracy": -1.0, "mean_target_support_size": 5.912109375 }, { "model": "cyclic_transition_ladder", "variant": "pilot_primitive_router", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "primitive_router", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 500, "length": 3, "query_type": "B", "k": 9, "n": 512, "decoder_query_target_mass": 0.9999813437461853, "decoder_query_target_nll": 1.2284263372421265, "decoder_query_top1_on_support": 1.0, "mean_decoder_query_support_size": 5.005859375, "decoder_belief_target_mass": 0.9999337196350098, "decoder_belief_target_nll": 1.5004963874816895, "decoder_belief_top1_on_support": 1.0, "mean_decoder_belief_support_size": 5.912109375, "mean_slot_purity": 0.7600073218345642, "mean_slot_entropy": 0.49823057651519775, "mean_weight_entropy": 1.5004939436912537, "mean_route_entropy": -1.0, "mean_route_accuracy": -1.0, "mean_target_support_size": 5.912109375 }, { "model": "cyclic_transition_ladder", "variant": "pilot_primitive_router", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "primitive_router", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 500, "length": 3, "query_type": "B", "k": 12, "n": 512, "decoder_query_target_mass": 0.9999813437461853, "decoder_query_target_nll": 1.2284263372421265, "decoder_query_top1_on_support": 1.0, "mean_decoder_query_support_size": 5.005859375, "decoder_belief_target_mass": 0.9999337196350098, "decoder_belief_target_nll": 1.5004963874816895, "decoder_belief_top1_on_support": 1.0, "mean_decoder_belief_support_size": 5.912109375, "mean_slot_purity": 0.7600073218345642, "mean_slot_entropy": 0.49823057651519775, "mean_weight_entropy": 1.5004939436912537, "mean_route_entropy": -1.0, "mean_route_accuracy": -1.0, "mean_target_support_size": 5.912109375 }, { "model": "cyclic_transition_ladder", "variant": "pilot_primitive_router", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "primitive_router", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 500, "length": 3, "query_type": "A_PLUS_B", "k": 0, "n": 512, "decoder_query_target_mass": 0.5324928760528564, "decoder_query_target_nll": 2.397895336151123, "decoder_query_top1_on_support": 0.513671875, "mean_decoder_query_support_size": 5.857421875, "decoder_belief_target_mass": 0.06409797258675098, "decoder_belief_target_nll": 15.616546154022217, "decoder_belief_top1_on_support": 0.09765625, "mean_decoder_belief_support_size": 6.25390625, "mean_slot_purity": 0.999997615814209, "mean_slot_entropy": 3.839540295302868e-05, "mean_weight_entropy": 2.397895336151123, "mean_route_entropy": -1.0, "mean_route_accuracy": -1.0, "mean_target_support_size": 6.25390625 }, { "model": "cyclic_transition_ladder", "variant": "pilot_primitive_router", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "primitive_router", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 500, "length": 3, "query_type": "A_PLUS_B", "k": 1, "n": 512, "decoder_query_target_mass": 0.5569657385349274, "decoder_query_target_nll": 4.292019605636597, "decoder_query_top1_on_support": 0.5546875, "mean_decoder_query_support_size": 5.857421875, "decoder_belief_target_mass": 0.164533369243145, "decoder_belief_target_nll": 13.394010066986084, "decoder_belief_top1_on_support": 0.1875, "mean_decoder_belief_support_size": 6.25390625, "mean_slot_purity": 0.8769612312316895, "mean_slot_entropy": 0.19487348943948746, "mean_weight_entropy": 2.0218923091888428, "mean_route_entropy": 0.0003715308557730168, "mean_route_accuracy": 1.0, "mean_target_support_size": 6.25390625 }, { "model": "cyclic_transition_ladder", "variant": "pilot_primitive_router", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "primitive_router", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 500, "length": 3, "query_type": "A_PLUS_B", "k": 2, "n": 512, "decoder_query_target_mass": 0.6029802560806274, "decoder_query_target_nll": 5.1051154136657715, "decoder_query_top1_on_support": 0.615234375, "mean_decoder_query_support_size": 5.857421875, "decoder_belief_target_mass": 0.2755606174468994, "decoder_belief_target_nll": 10.559035778045654, "decoder_belief_top1_on_support": 0.31640625, "mean_decoder_belief_support_size": 6.25390625, "mean_slot_purity": 0.8093588352203369, "mean_slot_entropy": 0.35972653329372406, "mean_weight_entropy": 1.7280826568603516, "mean_route_entropy": 0.0003229006542824209, "mean_route_accuracy": 1.0, "mean_target_support_size": 6.25390625 }, { "model": "cyclic_transition_ladder", "variant": "pilot_primitive_router", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "primitive_router", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 500, "length": 3, "query_type": "A_PLUS_B", "k": 3, "n": 512, "decoder_query_target_mass": 0.9999689757823944, "decoder_query_target_nll": 1.4522852301597595, "decoder_query_top1_on_support": 1.0, "mean_decoder_query_support_size": 5.857421875, "decoder_belief_target_mass": 0.9999352693557739, "decoder_belief_target_nll": 1.5494862794876099, "decoder_belief_top1_on_support": 1.0, "mean_decoder_belief_support_size": 6.25390625, "mean_slot_purity": 0.7717578411102295, "mean_slot_entropy": 0.47104403376579285, "mean_weight_entropy": 1.5494816303253174, "mean_route_entropy": 0.000347330336808227, "mean_route_accuracy": 1.0, "mean_target_support_size": 6.25390625 }, { "model": "cyclic_transition_ladder", "variant": "pilot_primitive_router", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "primitive_router", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 500, "length": 3, "query_type": "A_PLUS_B", "k": 6, "n": 512, "decoder_query_target_mass": 0.9999690055847168, "decoder_query_target_nll": 1.4522851705551147, "decoder_query_top1_on_support": 1.0, "mean_decoder_query_support_size": 5.857421875, "decoder_belief_target_mass": 0.9999352693557739, "decoder_belief_target_nll": 1.5494862794876099, "decoder_belief_top1_on_support": 1.0, "mean_decoder_belief_support_size": 6.25390625, "mean_slot_purity": 0.7717578411102295, "mean_slot_entropy": 0.47104403376579285, "mean_weight_entropy": 1.5494816303253174, "mean_route_entropy": -1.0, "mean_route_accuracy": -1.0, "mean_target_support_size": 6.25390625 }, { "model": "cyclic_transition_ladder", "variant": "pilot_primitive_router", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "primitive_router", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 500, "length": 3, "query_type": "A_PLUS_B", "k": 9, "n": 512, "decoder_query_target_mass": 0.9999690055847168, "decoder_query_target_nll": 1.4522851705551147, "decoder_query_top1_on_support": 1.0, "mean_decoder_query_support_size": 5.857421875, "decoder_belief_target_mass": 0.9999352693557739, "decoder_belief_target_nll": 1.5494862794876099, "decoder_belief_top1_on_support": 1.0, "mean_decoder_belief_support_size": 6.25390625, "mean_slot_purity": 0.7717578411102295, "mean_slot_entropy": 0.47104403376579285, "mean_weight_entropy": 1.5494816303253174, "mean_route_entropy": -1.0, "mean_route_accuracy": -1.0, "mean_target_support_size": 6.25390625 }, { "model": "cyclic_transition_ladder", "variant": "pilot_primitive_router", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "primitive_router", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 500, "length": 3, "query_type": "A_PLUS_B", "k": 12, "n": 512, "decoder_query_target_mass": 0.9999690055847168, "decoder_query_target_nll": 1.4522851705551147, "decoder_query_top1_on_support": 1.0, "mean_decoder_query_support_size": 5.857421875, "decoder_belief_target_mass": 0.9999352693557739, "decoder_belief_target_nll": 1.5494862794876099, "decoder_belief_top1_on_support": 1.0, "mean_decoder_belief_support_size": 6.25390625, "mean_slot_purity": 0.7717578411102295, "mean_slot_entropy": 0.47104403376579285, "mean_weight_entropy": 1.5494816303253174, "mean_route_entropy": -1.0, "mean_route_accuracy": -1.0, "mean_target_support_size": 6.25390625 }, { "model": "cyclic_transition_ladder", "variant": "pilot_primitive_router", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "primitive_router", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 500, "length": 3, "query_type": "A_MINUS_B", "k": 0, "n": 512, "decoder_query_target_mass": 0.47265617549419403, "decoder_query_target_nll": 13.3654146194458, "decoder_query_top1_on_support": 0.47265625, "mean_decoder_query_support_size": 4.90625, "decoder_belief_target_mass": 0.06409796886146069, "decoder_belief_target_nll": 15.763309478759766, "decoder_belief_top1_on_support": 0.107421875, "mean_decoder_belief_support_size": 6.041015625, "mean_slot_purity": 0.999997615814209, "mean_slot_entropy": 3.839540295302868e-05, "mean_weight_entropy": 2.397895336151123, "mean_route_entropy": -1.0, "mean_route_accuracy": -1.0, "mean_target_support_size": 6.041015625 }, { "model": "cyclic_transition_ladder", "variant": "pilot_primitive_router", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "primitive_router", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 500, "length": 3, "query_type": "A_MINUS_B", "k": 1, "n": 512, "decoder_query_target_mass": 0.5110073983669281, "decoder_query_target_nll": 7.933083534240723, "decoder_query_top1_on_support": 0.525390625, "mean_decoder_query_support_size": 4.90625, "decoder_belief_target_mass": 0.1726420596241951, "decoder_belief_target_nll": 13.21288251876831, "decoder_belief_top1_on_support": 0.2109375, "mean_decoder_belief_support_size": 6.041015625, "mean_slot_purity": 0.8690019249916077, "mean_slot_entropy": 0.20985938608646393, "mean_weight_entropy": 2.011256158351898, "mean_route_entropy": 0.00034195627085864544, "mean_route_accuracy": 1.0, "mean_target_support_size": 6.041015625 }, { "model": "cyclic_transition_ladder", "variant": "pilot_primitive_router", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "primitive_router", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 500, "length": 3, "query_type": "A_MINUS_B", "k": 2, "n": 512, "decoder_query_target_mass": 0.5781202912330627, "decoder_query_target_nll": 6.190404415130615, "decoder_query_top1_on_support": 0.5859375, "mean_decoder_query_support_size": 4.90625, "decoder_belief_target_mass": 0.3002113103866577, "decoder_belief_target_nll": 10.161686897277832, "decoder_belief_top1_on_support": 0.345703125, "mean_decoder_belief_support_size": 6.041015625, "mean_slot_purity": 0.7959901988506317, "mean_slot_entropy": 0.3907417058944702, "mean_weight_entropy": 1.7201481461524963, "mean_route_entropy": 0.0003217084158677608, "mean_route_accuracy": 1.0, "mean_target_support_size": 6.041015625 }, { "model": "cyclic_transition_ladder", "variant": "pilot_primitive_router", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "primitive_router", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 500, "length": 3, "query_type": "A_MINUS_B", "k": 3, "n": 512, "decoder_query_target_mass": 0.9999630153179169, "decoder_query_target_nll": 1.143086016178131, "decoder_query_top1_on_support": 1.0, "mean_decoder_query_support_size": 4.90625, "decoder_belief_target_mass": 0.9999370872974396, "decoder_belief_target_nll": 1.5218163132667542, "decoder_belief_top1_on_support": 1.0, "mean_decoder_belief_support_size": 6.041015625, "mean_slot_purity": 0.7537124156951904, "mean_slot_entropy": 0.5237121731042862, "mean_weight_entropy": 1.5218117833137512, "mean_route_entropy": 0.0003492558898869902, "mean_route_accuracy": 1.0, "mean_target_support_size": 6.041015625 }, { "model": "cyclic_transition_ladder", "variant": "pilot_primitive_router", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "primitive_router", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 500, "length": 3, "query_type": "A_MINUS_B", "k": 6, "n": 512, "decoder_query_target_mass": 0.9999629855155945, "decoder_query_target_nll": 1.143086016178131, "decoder_query_top1_on_support": 1.0, "mean_decoder_query_support_size": 4.90625, "decoder_belief_target_mass": 0.9999370872974396, "decoder_belief_target_nll": 1.5218163132667542, "decoder_belief_top1_on_support": 1.0, "mean_decoder_belief_support_size": 6.041015625, "mean_slot_purity": 0.7537124156951904, "mean_slot_entropy": 0.5237121731042862, "mean_weight_entropy": 1.5218117833137512, "mean_route_entropy": -1.0, "mean_route_accuracy": -1.0, "mean_target_support_size": 6.041015625 }, { "model": "cyclic_transition_ladder", "variant": "pilot_primitive_router", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "primitive_router", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 500, "length": 3, "query_type": "A_MINUS_B", "k": 9, "n": 512, "decoder_query_target_mass": 0.9999629855155945, "decoder_query_target_nll": 1.143086016178131, "decoder_query_top1_on_support": 1.0, "mean_decoder_query_support_size": 4.90625, "decoder_belief_target_mass": 0.9999370872974396, "decoder_belief_target_nll": 1.5218163132667542, "decoder_belief_top1_on_support": 1.0, "mean_decoder_belief_support_size": 6.041015625, "mean_slot_purity": 0.7537124156951904, "mean_slot_entropy": 0.5237121731042862, "mean_weight_entropy": 1.5218117833137512, "mean_route_entropy": -1.0, "mean_route_accuracy": -1.0, "mean_target_support_size": 6.041015625 }, { "model": "cyclic_transition_ladder", "variant": "pilot_primitive_router", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "primitive_router", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 500, "length": 3, "query_type": "A_MINUS_B", "k": 12, "n": 512, "decoder_query_target_mass": 0.9999630153179169, "decoder_query_target_nll": 1.143086016178131, "decoder_query_top1_on_support": 1.0, "mean_decoder_query_support_size": 4.90625, "decoder_belief_target_mass": 0.9999370872974396, "decoder_belief_target_nll": 1.5218163132667542, "decoder_belief_top1_on_support": 1.0, "mean_decoder_belief_support_size": 6.041015625, "mean_slot_purity": 0.7537124156951904, "mean_slot_entropy": 0.5237121731042862, "mean_weight_entropy": 1.5218117833137512, "mean_route_entropy": -1.0, "mean_route_accuracy": -1.0, "mean_target_support_size": 6.041015625 }, { "model": "cyclic_transition_ladder", "variant": "pilot_primitive_router", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "primitive_router", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 500, "length": 6, "query_type": "A", "k": 0, "n": 512, "decoder_query_target_mass": 0.2736150622367859, "decoder_query_target_nll": 2.397895336151123, "decoder_query_top1_on_support": 0.279296875, "mean_decoder_query_support_size": 3.009765625, "decoder_belief_target_mass": 0.030894880183041096, "decoder_belief_target_nll": 16.201786994934082, "decoder_belief_top1_on_support": 0.041015625, "mean_decoder_belief_support_size": 3.615234375, "mean_slot_purity": 0.999997615814209, "mean_slot_entropy": 3.839540295302868e-05, "mean_weight_entropy": 2.397895336151123, "mean_route_entropy": -1.0, "mean_route_accuracy": -1.0, "mean_target_support_size": 3.615234375 }, { "model": "cyclic_transition_ladder", "variant": "pilot_primitive_router", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "primitive_router", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 500, "length": 6, "query_type": "A", "k": 1, "n": 512, "decoder_query_target_mass": 0.30243799090385437, "decoder_query_target_nll": 5.863611459732056, "decoder_query_top1_on_support": 0.31640625, "mean_decoder_query_support_size": 3.009765625, "decoder_belief_target_mass": 0.03900293447077274, "decoder_belief_target_nll": 15.879072666168213, "decoder_belief_top1_on_support": 0.0390625, "mean_decoder_belief_support_size": 3.615234375, "mean_slot_purity": 0.8614259362220764, "mean_slot_entropy": 0.22234636545181274, "mean_weight_entropy": 1.9909549951553345, "mean_route_entropy": 0.00033543395693413913, "mean_route_accuracy": 1.0, "mean_target_support_size": 3.615234375 }, { "model": "cyclic_transition_ladder", "variant": "pilot_primitive_router", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "primitive_router", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 500, "length": 6, "query_type": "A", "k": 2, "n": 512, "decoder_query_target_mass": 0.32821255922317505, "decoder_query_target_nll": 7.162192106246948, "decoder_query_top1_on_support": 0.310546875, "mean_decoder_query_support_size": 3.009765625, "decoder_belief_target_mass": 0.054035669192671776, "decoder_belief_target_nll": 15.648300647735596, "decoder_belief_top1_on_support": 0.06640625, "mean_decoder_belief_support_size": 3.615234375, "mean_slot_purity": 0.7849280834197998, "mean_slot_entropy": 0.4076748639345169, "mean_weight_entropy": 1.6638678312301636, "mean_route_entropy": 0.00032480225490871817, "mean_route_accuracy": 1.0, "mean_target_support_size": 3.615234375 }, { "model": "cyclic_transition_ladder", "variant": "pilot_primitive_router", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "primitive_router", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 500, "length": 6, "query_type": "A", "k": 3, "n": 512, "decoder_query_target_mass": 0.3798501789569855, "decoder_query_target_nll": 7.0194361209869385, "decoder_query_top1_on_support": 0.3828125, "mean_decoder_query_support_size": 3.009765625, "decoder_belief_target_mass": 0.07270804233849049, "decoder_belief_target_nll": 15.076384544372559, "decoder_belief_top1_on_support": 0.083984375, "mean_decoder_belief_support_size": 3.615234375, "mean_slot_purity": 0.7437974512577057, "mean_slot_entropy": 0.5438433885574341, "mean_weight_entropy": 1.4797999262809753, "mean_route_entropy": 0.0003347025776747614, "mean_route_accuracy": 1.0, "mean_target_support_size": 3.615234375 }, { "model": "cyclic_transition_ladder", "variant": "pilot_primitive_router", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "primitive_router", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 500, "length": 6, "query_type": "A", "k": 6, "n": 512, "decoder_query_target_mass": 0.9999405741691589, "decoder_query_target_nll": 0.7840341627597809, "decoder_query_top1_on_support": 1.0, "mean_decoder_query_support_size": 3.009765625, "decoder_belief_target_mass": 0.9998491406440735, "decoder_belief_target_nll": 0.9942581653594971, "decoder_belief_top1_on_support": 1.0, "mean_decoder_belief_support_size": 3.615234375, "mean_slot_purity": 0.6582612991333008, "mean_slot_entropy": 0.8695366382598877, "mean_weight_entropy": 0.9943209886550903, "mean_route_entropy": 0.0003562292404239997, "mean_route_accuracy": 1.0, "mean_target_support_size": 3.615234375 }, { "model": "cyclic_transition_ladder", "variant": "pilot_primitive_router", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "primitive_router", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 500, "length": 6, "query_type": "A", "k": 9, "n": 512, "decoder_query_target_mass": 0.9999405741691589, "decoder_query_target_nll": 0.7840341627597809, "decoder_query_top1_on_support": 1.0, "mean_decoder_query_support_size": 3.009765625, "decoder_belief_target_mass": 0.9998491406440735, "decoder_belief_target_nll": 0.9942581653594971, "decoder_belief_top1_on_support": 1.0, "mean_decoder_belief_support_size": 3.615234375, "mean_slot_purity": 0.6582612991333008, "mean_slot_entropy": 0.8695366382598877, "mean_weight_entropy": 0.9943209886550903, "mean_route_entropy": -1.0, "mean_route_accuracy": -1.0, "mean_target_support_size": 3.615234375 }, { "model": "cyclic_transition_ladder", "variant": "pilot_primitive_router", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "primitive_router", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 500, "length": 6, "query_type": "A", "k": 12, "n": 512, "decoder_query_target_mass": 0.9999405741691589, "decoder_query_target_nll": 0.7840341627597809, "decoder_query_top1_on_support": 1.0, "mean_decoder_query_support_size": 3.009765625, "decoder_belief_target_mass": 0.9998491406440735, "decoder_belief_target_nll": 0.9942581653594971, "decoder_belief_top1_on_support": 1.0, "mean_decoder_belief_support_size": 3.615234375, "mean_slot_purity": 0.6582612991333008, "mean_slot_entropy": 0.8695366382598877, "mean_weight_entropy": 0.9943209886550903, "mean_route_entropy": -1.0, "mean_route_accuracy": -1.0, "mean_target_support_size": 3.615234375 }, { "model": "cyclic_transition_ladder", "variant": "pilot_primitive_router", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "primitive_router", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 500, "length": 6, "query_type": "B", "k": 0, "n": 512, "decoder_query_target_mass": 0.29367896914482117, "decoder_query_target_nll": 2.397895336151123, "decoder_query_top1_on_support": 0.291015625, "mean_decoder_query_support_size": 3.23046875, "decoder_belief_target_mass": 0.031427549198269844, "decoder_belief_target_nll": 16.20088005065918, "decoder_belief_top1_on_support": 0.025390625, "mean_decoder_belief_support_size": 3.689453125, "mean_slot_purity": 0.999997615814209, "mean_slot_entropy": 3.839540295302868e-05, "mean_weight_entropy": 2.397895336151123, "mean_route_entropy": -1.0, "mean_route_accuracy": -1.0, "mean_target_support_size": 3.689453125 }, { "model": "cyclic_transition_ladder", "variant": "pilot_primitive_router", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "primitive_router", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 500, "length": 6, "query_type": "B", "k": 1, "n": 512, "decoder_query_target_mass": 0.3156364858150482, "decoder_query_target_nll": 5.5915207862854, "decoder_query_top1_on_support": 0.328125, "mean_decoder_query_support_size": 3.23046875, "decoder_belief_target_mass": 0.04211066663265228, "decoder_belief_target_nll": 15.6771821975708, "decoder_belief_top1_on_support": 0.0546875, "mean_decoder_belief_support_size": 3.689453125, "mean_slot_purity": 0.8646510541439056, "mean_slot_entropy": 0.2151549756526947, "mean_weight_entropy": 1.9889029264450073, "mean_route_entropy": 0.00034531652636360377, "mean_route_accuracy": 1.0, "mean_target_support_size": 3.689453125 }, { "model": "cyclic_transition_ladder", "variant": "pilot_primitive_router", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "primitive_router", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 500, "length": 6, "query_type": "B", "k": 2, "n": 512, "decoder_query_target_mass": 0.3621540516614914, "decoder_query_target_nll": 6.530924320220947, "decoder_query_top1_on_support": 0.359375, "mean_decoder_query_support_size": 3.23046875, "decoder_belief_target_mass": 0.0575278215110302, "decoder_belief_target_nll": 15.394655704498291, "decoder_belief_top1_on_support": 0.064453125, "mean_decoder_belief_support_size": 3.689453125, "mean_slot_purity": 0.8032482266426086, "mean_slot_entropy": 0.35748618841171265, "mean_weight_entropy": 1.736682415008545, "mean_route_entropy": 0.00035426541580818594, "mean_route_accuracy": 1.0, "mean_target_support_size": 3.689453125 }, { "model": "cyclic_transition_ladder", "variant": "pilot_primitive_router", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "primitive_router", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 500, "length": 6, "query_type": "B", "k": 3, "n": 512, "decoder_query_target_mass": 0.40112580358982086, "decoder_query_target_nll": 6.9802565574646, "decoder_query_top1_on_support": 0.384765625, "mean_decoder_query_support_size": 3.23046875, "decoder_belief_target_mass": 0.08152558654546738, "decoder_belief_target_nll": 14.518213272094727, "decoder_belief_top1_on_support": 0.08984375, "mean_decoder_belief_support_size": 3.689453125, "mean_slot_purity": 0.7578997313976288, "mean_slot_entropy": 0.49595509469509125, "mean_weight_entropy": 1.5041415691375732, "mean_route_entropy": 0.00034989840060006827, "mean_route_accuracy": 1.0, "mean_target_support_size": 3.689453125 }, { "model": "cyclic_transition_ladder", "variant": "pilot_primitive_router", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "primitive_router", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 500, "length": 6, "query_type": "B", "k": 6, "n": 512, "decoder_query_target_mass": 0.9999475479125977, "decoder_query_target_nll": 0.841717928647995, "decoder_query_top1_on_support": 1.0, "mean_decoder_query_support_size": 3.23046875, "decoder_belief_target_mass": 0.9998529255390167, "decoder_belief_target_nll": 1.0146399438381195, "decoder_belief_top1_on_support": 1.0, "mean_decoder_belief_support_size": 3.689453125, "mean_slot_purity": 0.6683574616909027, "mean_slot_entropy": 0.8183096349239349, "mean_weight_entropy": 1.0147209167480469, "mean_route_entropy": 0.0003407969052204862, "mean_route_accuracy": 1.0, "mean_target_support_size": 3.689453125 }, { "model": "cyclic_transition_ladder", "variant": "pilot_primitive_router", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "primitive_router", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 500, "length": 6, "query_type": "B", "k": 9, "n": 512, "decoder_query_target_mass": 0.99994757771492, "decoder_query_target_nll": 0.841717928647995, "decoder_query_top1_on_support": 1.0, "mean_decoder_query_support_size": 3.23046875, "decoder_belief_target_mass": 0.9998529255390167, "decoder_belief_target_nll": 1.0146399438381195, "decoder_belief_top1_on_support": 1.0, "mean_decoder_belief_support_size": 3.689453125, "mean_slot_purity": 0.6683574616909027, "mean_slot_entropy": 0.8183096349239349, "mean_weight_entropy": 1.0147209167480469, "mean_route_entropy": -1.0, "mean_route_accuracy": -1.0, "mean_target_support_size": 3.689453125 }, { "model": "cyclic_transition_ladder", "variant": "pilot_primitive_router", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "primitive_router", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 500, "length": 6, "query_type": "B", "k": 12, "n": 512, "decoder_query_target_mass": 0.99994757771492, "decoder_query_target_nll": 0.841717928647995, "decoder_query_top1_on_support": 1.0, "mean_decoder_query_support_size": 3.23046875, "decoder_belief_target_mass": 0.9998529255390167, "decoder_belief_target_nll": 1.0146399438381195, "decoder_belief_top1_on_support": 1.0, "mean_decoder_belief_support_size": 3.689453125, "mean_slot_purity": 0.6683574616909027, "mean_slot_entropy": 0.8183096349239349, "mean_weight_entropy": 1.0147209167480469, "mean_route_entropy": -1.0, "mean_route_accuracy": -1.0, "mean_target_support_size": 3.689453125 }, { "model": "cyclic_transition_ladder", "variant": "pilot_primitive_router", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "primitive_router", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 500, "length": 6, "query_type": "A_PLUS_B", "k": 0, "n": 512, "decoder_query_target_mass": 0.2926136404275894, "decoder_query_target_nll": 2.397895336151123, "decoder_query_top1_on_support": 0.3046875, "mean_decoder_query_support_size": 3.21875, "decoder_belief_target_mass": 0.028409086167812347, "decoder_belief_target_nll": 16.214017868041992, "decoder_belief_top1_on_support": 0.037109375, "mean_decoder_belief_support_size": 3.38671875, "mean_slot_purity": 0.999997615814209, "mean_slot_entropy": 3.839540295302868e-05, "mean_weight_entropy": 2.397895336151123, "mean_route_entropy": -1.0, "mean_route_accuracy": -1.0, "mean_target_support_size": 3.38671875 }, { "model": "cyclic_transition_ladder", "variant": "pilot_primitive_router", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "primitive_router", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 500, "length": 6, "query_type": "A_PLUS_B", "k": 1, "n": 512, "decoder_query_target_mass": 0.2889440655708313, "decoder_query_target_nll": 5.205198287963867, "decoder_query_top1_on_support": 0.298828125, "mean_decoder_query_support_size": 3.21875, "decoder_belief_target_mass": 0.04681549221277237, "decoder_belief_target_nll": 15.682583332061768, "decoder_belief_top1_on_support": 0.05078125, "mean_decoder_belief_support_size": 3.38671875, "mean_slot_purity": 0.851394534111023, "mean_slot_entropy": 0.2369939237833023, "mean_weight_entropy": 1.9533761739730835, "mean_route_entropy": 0.0003403909067856148, "mean_route_accuracy": 1.0, "mean_target_support_size": 3.38671875 }, { "model": "cyclic_transition_ladder", "variant": "pilot_primitive_router", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "primitive_router", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 500, "length": 6, "query_type": "A_PLUS_B", "k": 2, "n": 512, "decoder_query_target_mass": 0.2870209217071533, "decoder_query_target_nll": 6.662882089614868, "decoder_query_top1_on_support": 0.291015625, "mean_decoder_query_support_size": 3.21875, "decoder_belief_target_mass": 0.05773475579917431, "decoder_belief_target_nll": 15.532008171081543, "decoder_belief_top1_on_support": 0.076171875, "mean_decoder_belief_support_size": 3.38671875, "mean_slot_purity": 0.7817320227622986, "mean_slot_entropy": 0.3989287167787552, "mean_weight_entropy": 1.6847256422042847, "mean_route_entropy": 0.0003419982094783336, "mean_route_accuracy": 1.0, "mean_target_support_size": 3.38671875 }, { "model": "cyclic_transition_ladder", "variant": "pilot_primitive_router", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "primitive_router", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 500, "length": 6, "query_type": "A_PLUS_B", "k": 3, "n": 512, "decoder_query_target_mass": 0.30353258550167084, "decoder_query_target_nll": 7.4065351486206055, "decoder_query_top1_on_support": 0.3125, "mean_decoder_query_support_size": 3.21875, "decoder_belief_target_mass": 0.09416177868843079, "decoder_belief_target_nll": 14.646194458007812, "decoder_belief_top1_on_support": 0.119140625, "mean_decoder_belief_support_size": 3.38671875, "mean_slot_purity": 0.7338176071643829, "mean_slot_entropy": 0.5500472784042358, "mean_weight_entropy": 1.4555519223213196, "mean_route_entropy": 0.0003443592431722209, "mean_route_accuracy": 1.0, "mean_target_support_size": 3.38671875 }, { "model": "cyclic_transition_ladder", "variant": "pilot_primitive_router", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "primitive_router", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 500, "length": 6, "query_type": "A_PLUS_B", "k": 6, "n": 512, "decoder_query_target_mass": 0.999889999628067, "decoder_query_target_nll": 0.8843976259231567, "decoder_query_top1_on_support": 1.0, "mean_decoder_query_support_size": 3.21875, "decoder_belief_target_mass": 0.9998548328876495, "decoder_belief_target_nll": 0.9564512372016907, "decoder_belief_top1_on_support": 1.0, "mean_decoder_belief_support_size": 3.38671875, "mean_slot_purity": 0.6350386440753937, "mean_slot_entropy": 0.9226740300655365, "mean_weight_entropy": 0.9565318822860718, "mean_route_entropy": 0.00034506036899983883, "mean_route_accuracy": 1.0, "mean_target_support_size": 3.38671875 }, { "model": "cyclic_transition_ladder", "variant": "pilot_primitive_router", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "primitive_router", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 500, "length": 6, "query_type": "A_PLUS_B", "k": 9, "n": 512, "decoder_query_target_mass": 0.999889999628067, "decoder_query_target_nll": 0.8843976259231567, "decoder_query_top1_on_support": 1.0, "mean_decoder_query_support_size": 3.21875, "decoder_belief_target_mass": 0.9998548328876495, "decoder_belief_target_nll": 0.9564512372016907, "decoder_belief_top1_on_support": 1.0, "mean_decoder_belief_support_size": 3.38671875, "mean_slot_purity": 0.6350386440753937, "mean_slot_entropy": 0.9226740300655365, "mean_weight_entropy": 0.9565318822860718, "mean_route_entropy": -1.0, "mean_route_accuracy": -1.0, "mean_target_support_size": 3.38671875 }, { "model": "cyclic_transition_ladder", "variant": "pilot_primitive_router", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "primitive_router", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 500, "length": 6, "query_type": "A_PLUS_B", "k": 12, "n": 512, "decoder_query_target_mass": 0.999889999628067, "decoder_query_target_nll": 0.8843976259231567, "decoder_query_top1_on_support": 1.0, "mean_decoder_query_support_size": 3.21875, "decoder_belief_target_mass": 0.9998548328876495, "decoder_belief_target_nll": 0.9564512372016907, "decoder_belief_top1_on_support": 1.0, "mean_decoder_belief_support_size": 3.38671875, "mean_slot_purity": 0.6350386440753937, "mean_slot_entropy": 0.9226740300655365, "mean_weight_entropy": 0.9565318822860718, "mean_route_entropy": -1.0, "mean_route_accuracy": -1.0, "mean_target_support_size": 3.38671875 }, { "model": "cyclic_transition_ladder", "variant": "pilot_primitive_router", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "primitive_router", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 500, "length": 6, "query_type": "A_MINUS_B", "k": 0, "n": 512, "decoder_query_target_mass": 0.3242186903953552, "decoder_query_target_nll": 13.719186305999756, "decoder_query_top1_on_support": 0.32421875, "mean_decoder_query_support_size": 3.408203125, "decoder_belief_target_mass": 0.03391334135085344, "decoder_belief_target_nll": 16.11708164215088, "decoder_belief_top1_on_support": 0.029296875, "mean_decoder_belief_support_size": 3.783203125, "mean_slot_purity": 0.999997615814209, "mean_slot_entropy": 3.839540295302868e-05, "mean_weight_entropy": 2.397895336151123, "mean_route_entropy": -1.0, "mean_route_accuracy": -1.0, "mean_target_support_size": 3.783203125 }, { "model": "cyclic_transition_ladder", "variant": "pilot_primitive_router", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "primitive_router", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 500, "length": 6, "query_type": "A_MINUS_B", "k": 1, "n": 512, "decoder_query_target_mass": 0.3224432170391083, "decoder_query_target_nll": 8.93432331085205, "decoder_query_top1_on_support": 0.328125, "mean_decoder_query_support_size": 3.408203125, "decoder_belief_target_mass": 0.04604608379304409, "decoder_belief_target_nll": 15.690613269805908, "decoder_belief_top1_on_support": 0.04296875, "mean_decoder_belief_support_size": 3.783203125, "mean_slot_purity": 0.8661015033721924, "mean_slot_entropy": 0.21338845789432526, "mean_weight_entropy": 1.9963540434837341, "mean_route_entropy": 0.00037236540811136365, "mean_route_accuracy": 1.0, "mean_target_support_size": 3.783203125 }, { "model": "cyclic_transition_ladder", "variant": "pilot_primitive_router", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "primitive_router", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 500, "length": 6, "query_type": "A_MINUS_B", "k": 2, "n": 512, "decoder_query_target_mass": 0.3105177730321884, "decoder_query_target_nll": 8.266419410705566, "decoder_query_top1_on_support": 0.30859375, "mean_decoder_query_support_size": 3.408203125, "decoder_belief_target_mass": 0.06510339863598347, "decoder_belief_target_nll": 15.197437286376953, "decoder_belief_top1_on_support": 0.0703125, "mean_decoder_belief_support_size": 3.783203125, "mean_slot_purity": 0.7958648502826691, "mean_slot_entropy": 0.375335693359375, "mean_weight_entropy": 1.6899656057357788, "mean_route_entropy": 0.0003613445005612448, "mean_route_accuracy": 1.0, "mean_target_support_size": 3.783203125 }, { "model": "cyclic_transition_ladder", "variant": "pilot_primitive_router", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "primitive_router", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 500, "length": 6, "query_type": "A_MINUS_B", "k": 3, "n": 512, "decoder_query_target_mass": 0.33750389516353607, "decoder_query_target_nll": 7.972452163696289, "decoder_query_top1_on_support": 0.3125, "mean_decoder_query_support_size": 3.408203125, "decoder_belief_target_mass": 0.09146890044212341, "decoder_belief_target_nll": 14.569271087646484, "decoder_belief_top1_on_support": 0.0859375, "mean_decoder_belief_support_size": 3.783203125, "mean_slot_purity": 0.7554861605167389, "mean_slot_entropy": 0.4937000274658203, "mean_weight_entropy": 1.4839302897453308, "mean_route_entropy": 0.0003329971223138273, "mean_route_accuracy": 1.0, "mean_target_support_size": 3.783203125 }, { "model": "cyclic_transition_ladder", "variant": "pilot_primitive_router", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "primitive_router", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 500, "length": 6, "query_type": "A_MINUS_B", "k": 6, "n": 512, "decoder_query_target_mass": 0.9998919665813446, "decoder_query_target_nll": 0.8671421408653259, "decoder_query_top1_on_support": 1.0, "mean_decoder_query_support_size": 3.408203125, "decoder_belief_target_mass": 0.999853253364563, "decoder_belief_target_nll": 1.0217082500457764, "decoder_belief_top1_on_support": 1.0, "mean_decoder_belief_support_size": 3.783203125, "mean_slot_purity": 0.6672395169734955, "mean_slot_entropy": 0.825376570224762, "mean_weight_entropy": 1.0217746496200562, "mean_route_entropy": 0.00034073837741743773, "mean_route_accuracy": 1.0, "mean_target_support_size": 3.783203125 }, { "model": "cyclic_transition_ladder", "variant": "pilot_primitive_router", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "primitive_router", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 500, "length": 6, "query_type": "A_MINUS_B", "k": 9, "n": 512, "decoder_query_target_mass": 0.9998919665813446, "decoder_query_target_nll": 0.8671421408653259, "decoder_query_top1_on_support": 1.0, "mean_decoder_query_support_size": 3.408203125, "decoder_belief_target_mass": 0.999853253364563, "decoder_belief_target_nll": 1.0217082500457764, "decoder_belief_top1_on_support": 1.0, "mean_decoder_belief_support_size": 3.783203125, "mean_slot_purity": 0.6672395169734955, "mean_slot_entropy": 0.825376570224762, "mean_weight_entropy": 1.0217746496200562, "mean_route_entropy": -1.0, "mean_route_accuracy": -1.0, "mean_target_support_size": 3.783203125 }, { "model": "cyclic_transition_ladder", "variant": "pilot_primitive_router", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "primitive_router", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 500, "length": 6, "query_type": "A_MINUS_B", "k": 12, "n": 512, "decoder_query_target_mass": 0.9998919665813446, "decoder_query_target_nll": 0.8671421408653259, "decoder_query_top1_on_support": 1.0, "mean_decoder_query_support_size": 3.408203125, "decoder_belief_target_mass": 0.999853253364563, "decoder_belief_target_nll": 1.0217082500457764, "decoder_belief_top1_on_support": 1.0, "mean_decoder_belief_support_size": 3.783203125, "mean_slot_purity": 0.6672395169734955, "mean_slot_entropy": 0.825376570224762, "mean_weight_entropy": 1.0217746496200562, "mean_route_entropy": -1.0, "mean_route_accuracy": -1.0, "mean_target_support_size": 3.783203125 }, { "model": "cyclic_transition_ladder", "variant": "pilot_primitive_router", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "primitive_router", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 500, "length": 9, "query_type": "A", "k": 0, "n": 512, "decoder_query_target_mass": 0.20436789095401764, "decoder_query_target_nll": 2.397895336151123, "decoder_query_top1_on_support": 0.20703125, "mean_decoder_query_support_size": 2.248046875, "decoder_belief_target_mass": 0.021661928854882717, "decoder_belief_target_nll": 16.15286636352539, "decoder_belief_top1_on_support": 0.025390625, "mean_decoder_belief_support_size": 2.623046875, "mean_slot_purity": 0.999997615814209, "mean_slot_entropy": 3.839540295302868e-05, "mean_weight_entropy": 2.397895336151123, "mean_route_entropy": -1.0, "mean_route_accuracy": -1.0, "mean_target_support_size": 2.623046875 }, { "model": "cyclic_transition_ladder", "variant": "pilot_primitive_router", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "primitive_router", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 500, "length": 9, "query_type": "A", "k": 1, "n": 512, "decoder_query_target_mass": 0.21602709591388702, "decoder_query_target_nll": 6.12308144569397, "decoder_query_top1_on_support": 0.23046875, "mean_decoder_query_support_size": 2.248046875, "decoder_belief_target_mass": 0.02846815064549446, "decoder_belief_target_nll": 15.939743041992188, "decoder_belief_top1_on_support": 0.037109375, "mean_decoder_belief_support_size": 2.623046875, "mean_slot_purity": 0.866692990064621, "mean_slot_entropy": 0.21158023923635483, "mean_weight_entropy": 1.993186354637146, "mean_route_entropy": 0.0003257015923736617, "mean_route_accuracy": 1.0, "mean_target_support_size": 2.623046875 }, { "model": "cyclic_transition_ladder", "variant": "pilot_primitive_router", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "primitive_router", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 500, "length": 9, "query_type": "A", "k": 2, "n": 512, "decoder_query_target_mass": 0.23206600546836853, "decoder_query_target_nll": 7.504704236984253, "decoder_query_top1_on_support": 0.234375, "mean_decoder_query_support_size": 2.248046875, "decoder_belief_target_mass": 0.03302537649869919, "decoder_belief_target_nll": 16.061121940612793, "decoder_belief_top1_on_support": 0.03515625, "mean_decoder_belief_support_size": 2.623046875, "mean_slot_purity": 0.7942158579826355, "mean_slot_entropy": 0.37694379687309265, "mean_weight_entropy": 1.6975546479225159, "mean_route_entropy": 0.000339335500029847, "mean_route_accuracy": 1.0, "mean_target_support_size": 2.623046875 }, { "model": "cyclic_transition_ladder", "variant": "pilot_primitive_router", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "primitive_router", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 500, "length": 9, "query_type": "A", "k": 3, "n": 512, "decoder_query_target_mass": 0.23887287080287933, "decoder_query_target_nll": 8.25239372253418, "decoder_query_top1_on_support": 0.265625, "mean_decoder_query_support_size": 2.248046875, "decoder_belief_target_mass": 0.04083767905831337, "decoder_belief_target_nll": 16.12129831314087, "decoder_belief_top1_on_support": 0.046875, "mean_decoder_belief_support_size": 2.623046875, "mean_slot_purity": 0.7451487481594086, "mean_slot_entropy": 0.525376707315445, "mean_weight_entropy": 1.4352070689201355, "mean_route_entropy": 0.0003551437985152006, "mean_route_accuracy": 1.0, "mean_target_support_size": 2.623046875 }, { "model": "cyclic_transition_ladder", "variant": "pilot_primitive_router", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "primitive_router", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 500, "length": 9, "query_type": "A", "k": 6, "n": 512, "decoder_query_target_mass": 0.3600146323442459, "decoder_query_target_nll": 8.152488231658936, "decoder_query_top1_on_support": 0.380859375, "mean_decoder_query_support_size": 2.248046875, "decoder_belief_target_mass": 0.12019729241728783, "decoder_belief_target_nll": 14.26258373260498, "decoder_belief_top1_on_support": 0.140625, "mean_decoder_belief_support_size": 2.623046875, "mean_slot_purity": 0.6595898270606995, "mean_slot_entropy": 0.8542106747627258, "mean_weight_entropy": 0.9813703894615173, "mean_route_entropy": 0.0003559127217158675, "mean_route_accuracy": 1.0, "mean_target_support_size": 2.623046875 }, { "model": "cyclic_transition_ladder", "variant": "pilot_primitive_router", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "primitive_router", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 500, "length": 9, "query_type": "A", "k": 9, "n": 512, "decoder_query_target_mass": 0.9998950958251953, "decoder_query_target_nll": 0.5430870652198792, "decoder_query_top1_on_support": 1.0, "mean_decoder_query_support_size": 2.248046875, "decoder_belief_target_mass": 0.9997605085372925, "decoder_belief_target_nll": 0.682287335395813, "decoder_belief_top1_on_support": 1.0, "mean_decoder_belief_support_size": 2.623046875, "mean_slot_purity": 0.5910454094409943, "mean_slot_entropy": 1.1097664833068848, "mean_weight_entropy": 0.6824472844600677, "mean_route_entropy": 0.0003387501637917012, "mean_route_accuracy": 1.0, "mean_target_support_size": 2.623046875 }, { "model": "cyclic_transition_ladder", "variant": "pilot_primitive_router", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "primitive_router", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 500, "length": 9, "query_type": "A", "k": 12, "n": 512, "decoder_query_target_mass": 0.9998950958251953, "decoder_query_target_nll": 0.5430870652198792, "decoder_query_top1_on_support": 1.0, "mean_decoder_query_support_size": 2.248046875, "decoder_belief_target_mass": 0.9997605085372925, "decoder_belief_target_nll": 0.682287335395813, "decoder_belief_top1_on_support": 1.0, "mean_decoder_belief_support_size": 2.623046875, "mean_slot_purity": 0.5910454094409943, "mean_slot_entropy": 1.1097664833068848, "mean_weight_entropy": 0.6824472844600677, "mean_route_entropy": -1.0, "mean_route_accuracy": -1.0, "mean_target_support_size": 2.623046875 }, { "model": "cyclic_transition_ladder", "variant": "pilot_primitive_router", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "primitive_router", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 500, "length": 9, "query_type": "B", "k": 0, "n": 512, "decoder_query_target_mass": 0.19442471116781235, "decoder_query_target_nll": 2.397895336151123, "decoder_query_top1_on_support": 0.1796875, "mean_decoder_query_support_size": 2.138671875, "decoder_belief_target_mass": 0.02024147380143404, "decoder_belief_target_nll": 16.4876127243042, "decoder_belief_top1_on_support": 0.02734375, "mean_decoder_belief_support_size": 2.40234375, "mean_slot_purity": 0.999997615814209, "mean_slot_entropy": 3.839540295302868e-05, "mean_weight_entropy": 2.397895336151123, "mean_route_entropy": -1.0, "mean_route_accuracy": -1.0, "mean_target_support_size": 2.40234375 }, { "model": "cyclic_transition_ladder", "variant": "pilot_primitive_router", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "primitive_router", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 500, "length": 9, "query_type": "B", "k": 1, "n": 512, "decoder_query_target_mass": 0.20362820476293564, "decoder_query_target_nll": 6.47435450553894, "decoder_query_top1_on_support": 0.197265625, "mean_decoder_query_support_size": 2.138671875, "decoder_belief_target_mass": 0.02178025431931019, "decoder_belief_target_nll": 16.29813575744629, "decoder_belief_top1_on_support": 0.029296875, "mean_decoder_belief_support_size": 2.40234375, "mean_slot_purity": 0.8506541550159454, "mean_slot_entropy": 0.2392643317580223, "mean_weight_entropy": 1.9573357105255127, "mean_route_entropy": 0.00033857817470561713, "mean_route_accuracy": 1.0, "mean_target_support_size": 2.40234375 }, { "model": "cyclic_transition_ladder", "variant": "pilot_primitive_router", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "primitive_router", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 500, "length": 9, "query_type": "B", "k": 2, "n": 512, "decoder_query_target_mass": 0.20729806274175644, "decoder_query_target_nll": 8.084302425384521, "decoder_query_top1_on_support": 0.216796875, "mean_decoder_query_support_size": 2.138671875, "decoder_belief_target_mass": 0.02278612833470106, "decoder_belief_target_nll": 16.43920135498047, "decoder_belief_top1_on_support": 0.025390625, "mean_decoder_belief_support_size": 2.40234375, "mean_slot_purity": 0.781259298324585, "mean_slot_entropy": 0.3974705636501312, "mean_weight_entropy": 1.6658741235733032, "mean_route_entropy": 0.00031498911266680807, "mean_route_accuracy": 1.0, "mean_target_support_size": 2.40234375 }, { "model": "cyclic_transition_ladder", "variant": "pilot_primitive_router", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "primitive_router", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 500, "length": 9, "query_type": "B", "k": 3, "n": 512, "decoder_query_target_mass": 0.21010896563529968, "decoder_query_target_nll": 8.795939445495605, "decoder_query_top1_on_support": 0.21875, "mean_decoder_query_support_size": 2.138671875, "decoder_belief_target_mass": 0.025952624157071114, "decoder_belief_target_nll": 16.490516662597656, "decoder_belief_top1_on_support": 0.02734375, "mean_decoder_belief_support_size": 2.40234375, "mean_slot_purity": 0.7331456243991852, "mean_slot_entropy": 0.5465466976165771, "mean_weight_entropy": 1.4422286748886108, "mean_route_entropy": 0.0003427817136980593, "mean_route_accuracy": 1.0, "mean_target_support_size": 2.40234375 }, { "model": "cyclic_transition_ladder", "variant": "pilot_primitive_router", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "primitive_router", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 500, "length": 9, "query_type": "B", "k": 6, "n": 512, "decoder_query_target_mass": 0.3573213666677475, "decoder_query_target_nll": 8.113773345947266, "decoder_query_top1_on_support": 0.341796875, "mean_decoder_query_support_size": 2.138671875, "decoder_belief_target_mass": 0.09702662378549576, "decoder_belief_target_nll": 14.866378784179688, "decoder_belief_top1_on_support": 0.109375, "mean_decoder_belief_support_size": 2.40234375, "mean_slot_purity": 0.6490113139152527, "mean_slot_entropy": 0.8594696819782257, "mean_weight_entropy": 0.9153112769126892, "mean_route_entropy": 0.0003438745334278792, "mean_route_accuracy": 1.0, "mean_target_support_size": 2.40234375 }, { "model": "cyclic_transition_ladder", "variant": "pilot_primitive_router", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "primitive_router", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 500, "length": 9, "query_type": "B", "k": 9, "n": 512, "decoder_query_target_mass": 0.9998832941055298, "decoder_query_target_nll": 0.4946826100349426, "decoder_query_top1_on_support": 1.0, "mean_decoder_query_support_size": 2.138671875, "decoder_belief_target_mass": 0.999743789434433, "decoder_belief_target_nll": 0.6081631779670715, "decoder_belief_top1_on_support": 1.0, "mean_decoder_belief_support_size": 2.40234375, "mean_slot_purity": 0.5819118916988373, "mean_slot_entropy": 1.1282269954681396, "mean_weight_entropy": 0.6083321869373322, "mean_route_entropy": 0.00034399474679958075, "mean_route_accuracy": 1.0, "mean_target_support_size": 2.40234375 }, { "model": "cyclic_transition_ladder", "variant": "pilot_primitive_router", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "primitive_router", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 500, "length": 9, "query_type": "B", "k": 12, "n": 512, "decoder_query_target_mass": 0.9998832941055298, "decoder_query_target_nll": 0.4946826100349426, "decoder_query_top1_on_support": 1.0, "mean_decoder_query_support_size": 2.138671875, "decoder_belief_target_mass": 0.999743789434433, "decoder_belief_target_nll": 0.6081631779670715, "decoder_belief_top1_on_support": 1.0, "mean_decoder_belief_support_size": 2.40234375, "mean_slot_purity": 0.5819118916988373, "mean_slot_entropy": 1.1282269954681396, "mean_weight_entropy": 0.6083321869373322, "mean_route_entropy": -1.0, "mean_route_accuracy": -1.0, "mean_target_support_size": 2.40234375 }, { "model": "cyclic_transition_ladder", "variant": "pilot_primitive_router", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "primitive_router", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 500, "length": 9, "query_type": "A_PLUS_B", "k": 0, "n": 512, "decoder_query_target_mass": 0.22105823457241058, "decoder_query_target_nll": 2.397895336151123, "decoder_query_top1_on_support": 0.234375, "mean_decoder_query_support_size": 2.431640625, "decoder_belief_target_mass": 0.024147716350853443, "decoder_belief_target_nll": 16.298270225524902, "decoder_belief_top1_on_support": 0.02734375, "mean_decoder_belief_support_size": 2.55078125, "mean_slot_purity": 0.999997615814209, "mean_slot_entropy": 3.839540295302868e-05, "mean_weight_entropy": 2.397895336151123, "mean_route_entropy": -1.0, "mean_route_accuracy": -1.0, "mean_target_support_size": 2.55078125 }, { "model": "cyclic_transition_ladder", "variant": "pilot_primitive_router", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "primitive_router", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 500, "length": 9, "query_type": "A_PLUS_B", "k": 1, "n": 512, "decoder_query_target_mass": 0.2253192812204361, "decoder_query_target_nll": 5.082322597503662, "decoder_query_top1_on_support": 0.201171875, "mean_decoder_query_support_size": 2.431640625, "decoder_belief_target_mass": 0.022431368939578533, "decoder_belief_target_nll": 16.330865383148193, "decoder_belief_top1_on_support": 0.01953125, "mean_decoder_belief_support_size": 2.55078125, "mean_slot_purity": 0.8529930710792542, "mean_slot_entropy": 0.23477014154195786, "mean_weight_entropy": 1.9600352048873901, "mean_route_entropy": 0.00034306192537769675, "mean_route_accuracy": 1.0, "mean_target_support_size": 2.55078125 }, { "model": "cyclic_transition_ladder", "variant": "pilot_primitive_router", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "primitive_router", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 500, "length": 9, "query_type": "A_PLUS_B", "k": 2, "n": 512, "decoder_query_target_mass": 0.22493459284305573, "decoder_query_target_nll": 6.421607255935669, "decoder_query_top1_on_support": 0.234375, "mean_decoder_query_support_size": 2.431640625, "decoder_belief_target_mass": 0.026455710642039776, "decoder_belief_target_nll": 16.49440574645996, "decoder_belief_top1_on_support": 0.033203125, "mean_decoder_belief_support_size": 2.55078125, "mean_slot_purity": 0.7880578339099884, "mean_slot_entropy": 0.3874709904193878, "mean_weight_entropy": 1.6983683705329895, "mean_route_entropy": 0.00035674295213539153, "mean_route_accuracy": 1.0, "mean_target_support_size": 2.55078125 }, { "model": "cyclic_transition_ladder", "variant": "pilot_primitive_router", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "primitive_router", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 500, "length": 9, "query_type": "A_PLUS_B", "k": 3, "n": 512, "decoder_query_target_mass": 0.2316821590065956, "decoder_query_target_nll": 7.3803629875183105, "decoder_query_top1_on_support": 0.24609375, "mean_decoder_query_support_size": 2.431640625, "decoder_belief_target_mass": 0.035954794846475124, "decoder_belief_target_nll": 16.365272521972656, "decoder_belief_top1_on_support": 0.03515625, "mean_decoder_belief_support_size": 2.55078125, "mean_slot_purity": 0.7428940534591675, "mean_slot_entropy": 0.528773158788681, "mean_weight_entropy": 1.4556968808174133, "mean_route_entropy": 0.0003223029198125005, "mean_route_accuracy": 1.0, "mean_target_support_size": 2.55078125 }, { "model": "cyclic_transition_ladder", "variant": "pilot_primitive_router", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "primitive_router", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 500, "length": 9, "query_type": "A_PLUS_B", "k": 6, "n": 512, "decoder_query_target_mass": 0.2121216282248497, "decoder_query_target_nll": 9.222660541534424, "decoder_query_top1_on_support": 0.21875, "mean_decoder_query_support_size": 2.431640625, "decoder_belief_target_mass": 0.07838542759418488, "decoder_belief_target_nll": 15.114310264587402, "decoder_belief_top1_on_support": 0.08203125, "mean_decoder_belief_support_size": 2.55078125, "mean_slot_purity": 0.6532246768474579, "mean_slot_entropy": 0.8530752062797546, "mean_weight_entropy": 0.971015214920044, "mean_route_entropy": 0.0003454697580309585, "mean_route_accuracy": 1.0, "mean_target_support_size": 2.55078125 }, { "model": "cyclic_transition_ladder", "variant": "pilot_primitive_router", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "primitive_router", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 500, "length": 9, "query_type": "A_PLUS_B", "k": 9, "n": 512, "decoder_query_target_mass": 0.9998071491718292, "decoder_query_target_nll": 0.6167428195476532, "decoder_query_top1_on_support": 1.0, "mean_decoder_query_support_size": 2.431640625, "decoder_belief_target_mass": 0.9997599720954895, "decoder_belief_target_nll": 0.6743583083152771, "decoder_belief_top1_on_support": 1.0, "mean_decoder_belief_support_size": 2.55078125, "mean_slot_purity": 0.5871718227863312, "mean_slot_entropy": 1.106706440448761, "mean_weight_entropy": 0.6745147407054901, "mean_route_entropy": 0.00037894988781772554, "mean_route_accuracy": 1.0, "mean_target_support_size": 2.55078125 }, { "model": "cyclic_transition_ladder", "variant": "pilot_primitive_router", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "primitive_router", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 500, "length": 9, "query_type": "A_PLUS_B", "k": 12, "n": 512, "decoder_query_target_mass": 0.9998071491718292, "decoder_query_target_nll": 0.6167428195476532, "decoder_query_top1_on_support": 1.0, "mean_decoder_query_support_size": 2.431640625, "decoder_belief_target_mass": 0.9997599720954895, "decoder_belief_target_nll": 0.6743583083152771, "decoder_belief_top1_on_support": 1.0, "mean_decoder_belief_support_size": 2.55078125, "mean_slot_purity": 0.5871718227863312, "mean_slot_entropy": 1.106706440448761, "mean_weight_entropy": 0.6745147407054901, "mean_route_entropy": -1.0, "mean_route_accuracy": -1.0, "mean_target_support_size": 2.55078125 }, { "model": "cyclic_transition_ladder", "variant": "pilot_primitive_router", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "primitive_router", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 500, "length": 9, "query_type": "A_MINUS_B", "k": 0, "n": 512, "decoder_query_target_mass": 0.2382812276482582, "decoder_query_target_nll": 13.897656917572021, "decoder_query_top1_on_support": 0.23828125, "mean_decoder_query_support_size": 2.56640625, "decoder_belief_target_mass": 0.025213060900568962, "decoder_belief_target_nll": 16.295551300048828, "decoder_belief_top1_on_support": 0.017578125, "mean_decoder_belief_support_size": 2.8125, "mean_slot_purity": 0.999997615814209, "mean_slot_entropy": 3.839540295302868e-05, "mean_weight_entropy": 2.397895336151123, "mean_route_entropy": -1.0, "mean_route_accuracy": -1.0, "mean_target_support_size": 2.8125 }, { "model": "cyclic_transition_ladder", "variant": "pilot_primitive_router", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "primitive_router", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 500, "length": 9, "query_type": "A_MINUS_B", "k": 1, "n": 512, "decoder_query_target_mass": 0.2420094907283783, "decoder_query_target_nll": 9.019840240478516, "decoder_query_top1_on_support": 0.2421875, "mean_decoder_query_support_size": 2.56640625, "decoder_belief_target_mass": 0.02778738271445036, "decoder_belief_target_nll": 16.209558486938477, "decoder_belief_top1_on_support": 0.02734375, "mean_decoder_belief_support_size": 2.8125, "mean_slot_purity": 0.8591175079345703, "mean_slot_entropy": 0.22406404465436935, "mean_weight_entropy": 1.972885251045227, "mean_route_entropy": 0.0003433021920500323, "mean_route_accuracy": 1.0, "mean_target_support_size": 2.8125 }, { "model": "cyclic_transition_ladder", "variant": "pilot_primitive_router", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "primitive_router", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 500, "length": 9, "query_type": "A_MINUS_B", "k": 2, "n": 512, "decoder_query_target_mass": 0.2307344228029251, "decoder_query_target_nll": 8.006723880767822, "decoder_query_top1_on_support": 0.224609375, "mean_decoder_query_support_size": 2.56640625, "decoder_belief_target_mass": 0.030361680313944817, "decoder_belief_target_nll": 16.19403076171875, "decoder_belief_top1_on_support": 0.033203125, "mean_decoder_belief_support_size": 2.8125, "mean_slot_purity": 0.8017300665378571, "mean_slot_entropy": 0.36094649136066437, "mean_weight_entropy": 1.7249545454978943, "mean_route_entropy": 0.00036935810931026936, "mean_route_accuracy": 1.0, "mean_target_support_size": 2.8125 }, { "model": "cyclic_transition_ladder", "variant": "pilot_primitive_router", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "primitive_router", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 500, "length": 9, "query_type": "A_MINUS_B", "k": 3, "n": 512, "decoder_query_target_mass": 0.22582225501537323, "decoder_query_target_nll": 8.047364711761475, "decoder_query_top1_on_support": 0.22265625, "mean_decoder_query_support_size": 2.56640625, "decoder_belief_target_mass": 0.036398179829120636, "decoder_belief_target_nll": 16.330371856689453, "decoder_belief_top1_on_support": 0.041015625, "mean_decoder_belief_support_size": 2.8125, "mean_slot_purity": 0.7597225606441498, "mean_slot_entropy": 0.48887430131435394, "mean_weight_entropy": 1.5260677337646484, "mean_route_entropy": 0.00034642769605852664, "mean_route_accuracy": 1.0, "mean_target_support_size": 2.8125 }, { "model": "cyclic_transition_ladder", "variant": "pilot_primitive_router", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "primitive_router", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 500, "length": 9, "query_type": "A_MINUS_B", "k": 6, "n": 512, "decoder_query_target_mass": 0.255205862224102, "decoder_query_target_nll": 8.54690408706665, "decoder_query_top1_on_support": 0.267578125, "mean_decoder_query_support_size": 2.56640625, "decoder_belief_target_mass": 0.09608104825019836, "decoder_belief_target_nll": 14.603612422943115, "decoder_belief_top1_on_support": 0.099609375, "mean_decoder_belief_support_size": 2.8125, "mean_slot_purity": 0.6742512285709381, "mean_slot_entropy": 0.7970572412014008, "mean_weight_entropy": 1.0227314233779907, "mean_route_entropy": 0.00034799468994606286, "mean_route_accuracy": 1.0, "mean_target_support_size": 2.8125 }, { "model": "cyclic_transition_ladder", "variant": "pilot_primitive_router", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "primitive_router", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 500, "length": 9, "query_type": "A_MINUS_B", "k": 9, "n": 512, "decoder_query_target_mass": 0.9998028576374054, "decoder_query_target_nll": 0.6542222499847412, "decoder_query_top1_on_support": 1.0, "mean_decoder_query_support_size": 2.56640625, "decoder_belief_target_mass": 0.999748945236206, "decoder_belief_target_nll": 0.7306901812553406, "decoder_belief_top1_on_support": 1.0, "mean_decoder_belief_support_size": 2.8125, "mean_slot_purity": 0.607384443283081, "mean_slot_entropy": 1.0458320379257202, "mean_weight_entropy": 0.7308489978313446, "mean_route_entropy": 0.0003722270776052028, "mean_route_accuracy": 1.0, "mean_target_support_size": 2.8125 }, { "model": "cyclic_transition_ladder", "variant": "pilot_primitive_router", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "primitive_router", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 500, "length": 9, "query_type": "A_MINUS_B", "k": 12, "n": 512, "decoder_query_target_mass": 0.9998028576374054, "decoder_query_target_nll": 0.6542222499847412, "decoder_query_top1_on_support": 1.0, "mean_decoder_query_support_size": 2.56640625, "decoder_belief_target_mass": 0.999748945236206, "decoder_belief_target_nll": 0.7306901812553406, "decoder_belief_top1_on_support": 1.0, "mean_decoder_belief_support_size": 2.8125, "mean_slot_purity": 0.607384443283081, "mean_slot_entropy": 1.0458320379257202, "mean_weight_entropy": 0.7308489978313446, "mean_route_entropy": -1.0, "mean_route_accuracy": -1.0, "mean_target_support_size": 2.8125 }, { "model": "cyclic_transition_ladder", "variant": "pilot_primitive_router", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "primitive_router", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 500, "length": 12, "query_type": "A", "k": 0, "n": 512, "decoder_query_target_mass": 0.16903408616781235, "decoder_query_target_nll": 2.397895336151123, "decoder_query_top1_on_support": 0.158203125, "mean_decoder_query_support_size": 1.859375, "decoder_belief_target_mass": 0.015269889030605555, "decoder_belief_target_nll": 16.432348251342773, "decoder_belief_top1_on_support": 0.01171875, "mean_decoder_belief_support_size": 2.046875, "mean_slot_purity": 0.999997615814209, "mean_slot_entropy": 3.839540295302868e-05, "mean_weight_entropy": 2.397895336151123, "mean_route_entropy": -1.0, "mean_route_accuracy": -1.0, "mean_target_support_size": 2.046875 }, { "model": "cyclic_transition_ladder", "variant": "pilot_primitive_router", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "primitive_router", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 500, "length": 12, "query_type": "A", "k": 1, "n": 512, "decoder_query_target_mass": 0.16122151911258698, "decoder_query_target_nll": 6.089410781860352, "decoder_query_top1_on_support": 0.154296875, "mean_decoder_query_support_size": 1.859375, "decoder_belief_target_mass": 0.01485554687678814, "decoder_belief_target_nll": 16.283698081970215, "decoder_belief_top1_on_support": 0.0234375, "mean_decoder_belief_support_size": 2.046875, "mean_slot_purity": 0.8754227161407471, "mean_slot_entropy": 0.1982385739684105, "mean_weight_entropy": 2.0225223302841187, "mean_route_entropy": 0.0003526811924530193, "mean_route_accuracy": 1.0, "mean_target_support_size": 2.046875 }, { "model": "cyclic_transition_ladder", "variant": "pilot_primitive_router", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "primitive_router", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 500, "length": 12, "query_type": "A", "k": 2, "n": 512, "decoder_query_target_mass": 0.1657792329788208, "decoder_query_target_nll": 7.867812156677246, "decoder_query_top1_on_support": 0.166015625, "mean_decoder_query_support_size": 1.859375, "decoder_belief_target_mass": 0.012369752395898104, "decoder_belief_target_nll": 16.792156219482422, "decoder_belief_top1_on_support": 0.01171875, "mean_decoder_belief_support_size": 2.046875, "mean_slot_purity": 0.8077841103076935, "mean_slot_entropy": 0.35508565604686737, "mean_weight_entropy": 1.7138166427612305, "mean_route_entropy": 0.0003489935479592532, "mean_route_accuracy": 1.0, "mean_target_support_size": 2.046875 }, { "model": "cyclic_transition_ladder", "variant": "pilot_primitive_router", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "primitive_router", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 500, "length": 12, "query_type": "A", "k": 3, "n": 512, "decoder_query_target_mass": 0.17690525203943253, "decoder_query_target_nll": 9.36870002746582, "decoder_query_top1_on_support": 0.16796875, "mean_decoder_query_support_size": 1.859375, "decoder_belief_target_mass": 0.022638294845819473, "decoder_belief_target_nll": 16.83303737640381, "decoder_belief_top1_on_support": 0.02734375, "mean_decoder_belief_support_size": 2.046875, "mean_slot_purity": 0.7584190666675568, "mean_slot_entropy": 0.5053583383560181, "mean_weight_entropy": 1.4545104503631592, "mean_route_entropy": 0.00035424603265710175, "mean_route_accuracy": 1.0, "mean_target_support_size": 2.046875 }, { "model": "cyclic_transition_ladder", "variant": "pilot_primitive_router", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "primitive_router", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 500, "length": 12, "query_type": "A", "k": 6, "n": 512, "decoder_query_target_mass": 0.22750411927700043, "decoder_query_target_nll": 9.439043998718262, "decoder_query_top1_on_support": 0.212890625, "mean_decoder_query_support_size": 1.859375, "decoder_belief_target_mass": 0.02976918499916792, "decoder_belief_target_nll": 16.73811149597168, "decoder_belief_top1_on_support": 0.02734375, "mean_decoder_belief_support_size": 2.046875, "mean_slot_purity": 0.6730387508869171, "mean_slot_entropy": 0.8164885342121124, "mean_weight_entropy": 1.0028997361660004, "mean_route_entropy": 0.0003372883074916899, "mean_route_accuracy": 1.0, "mean_target_support_size": 2.046875 }, { "model": "cyclic_transition_ladder", "variant": "pilot_primitive_router", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "primitive_router", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 500, "length": 12, "query_type": "A", "k": 9, "n": 512, "decoder_query_target_mass": 0.3750360310077667, "decoder_query_target_nll": 8.166024208068848, "decoder_query_top1_on_support": 0.384765625, "mean_decoder_query_support_size": 1.859375, "decoder_belief_target_mass": 0.08630935847759247, "decoder_belief_target_nll": 14.869909286499023, "decoder_belief_top1_on_support": 0.095703125, "mean_decoder_belief_support_size": 2.046875, "mean_slot_purity": 0.6019687652587891, "mean_slot_entropy": 1.0697612166404724, "mean_weight_entropy": 0.6817893087863922, "mean_route_entropy": 0.0003233681054553017, "mean_route_accuracy": 1.0, "mean_target_support_size": 2.046875 }, { "model": "cyclic_transition_ladder", "variant": "pilot_primitive_router", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "primitive_router", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 500, "length": 12, "query_type": "A", "k": 12, "n": 512, "decoder_query_target_mass": 0.9998307526111603, "decoder_query_target_nll": 0.3915087729692459, "decoder_query_top1_on_support": 1.0, "mean_decoder_query_support_size": 1.859375, "decoder_belief_target_mass": 0.999620109796524, "decoder_belief_target_nll": 0.4767141044139862, "decoder_belief_top1_on_support": 1.0, "mean_decoder_belief_support_size": 2.046875, "mean_slot_purity": 0.5434727072715759, "mean_slot_entropy": 1.2974021434783936, "mean_weight_entropy": 0.4768623262643814, "mean_route_entropy": 0.0003540814941516146, "mean_route_accuracy": 1.0, "mean_target_support_size": 2.046875 }, { "model": "cyclic_transition_ladder", "variant": "pilot_primitive_router", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "primitive_router", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 500, "length": 12, "query_type": "B", "k": 0, "n": 512, "decoder_query_target_mass": 0.15340909361839294, "decoder_query_target_nll": 2.397895336151123, "decoder_query_top1_on_support": 0.150390625, "mean_decoder_query_support_size": 1.6875, "decoder_belief_target_mass": 0.015802554320544004, "decoder_belief_target_nll": 16.169625282287598, "decoder_belief_top1_on_support": 0.01953125, "mean_decoder_belief_support_size": 1.869140625, "mean_slot_purity": 0.999997615814209, "mean_slot_entropy": 3.839540295302868e-05, "mean_weight_entropy": 2.397895336151123, "mean_route_entropy": -1.0, "mean_route_accuracy": -1.0, "mean_target_support_size": 1.869140625 }, { "model": "cyclic_transition_ladder", "variant": "pilot_primitive_router", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "primitive_router", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 500, "length": 12, "query_type": "B", "k": 1, "n": 512, "decoder_query_target_mass": 0.1522846519947052, "decoder_query_target_nll": 5.684526443481445, "decoder_query_top1_on_support": 0.16015625, "mean_decoder_query_support_size": 1.6875, "decoder_belief_target_mass": 0.017104651778936386, "decoder_belief_target_nll": 16.163784980773926, "decoder_belief_top1_on_support": 0.009765625, "mean_decoder_belief_support_size": 1.869140625, "mean_slot_purity": 0.8881170749664307, "mean_slot_entropy": 0.17544663697481155, "mean_weight_entropy": 2.0458465814590454, "mean_route_entropy": 0.00034985832462552935, "mean_route_accuracy": 1.0, "mean_target_support_size": 1.869140625 }, { "model": "cyclic_transition_ladder", "variant": "pilot_primitive_router", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "primitive_router", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 500, "length": 12, "query_type": "B", "k": 2, "n": 512, "decoder_query_target_mass": 0.1448274403810501, "decoder_query_target_nll": 7.376471996307373, "decoder_query_top1_on_support": 0.130859375, "mean_decoder_query_support_size": 1.6875, "decoder_belief_target_mass": 0.019915747456252575, "decoder_belief_target_nll": 16.234219551086426, "decoder_belief_top1_on_support": 0.0234375, "mean_decoder_belief_support_size": 1.869140625, "mean_slot_purity": 0.8246946334838867, "mean_slot_entropy": 0.32107000052928925, "mean_weight_entropy": 1.7562631368637085, "mean_route_entropy": 0.00034486112417653203, "mean_route_accuracy": 1.0, "mean_target_support_size": 1.869140625 }, { "model": "cyclic_transition_ladder", "variant": "pilot_primitive_router", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "primitive_router", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 500, "length": 12, "query_type": "B", "k": 3, "n": 512, "decoder_query_target_mass": 0.1645948514342308, "decoder_query_target_nll": 8.288332462310791, "decoder_query_top1_on_support": 0.16796875, "mean_decoder_query_support_size": 1.6875, "decoder_belief_target_mass": 0.020832940004765987, "decoder_belief_target_nll": 16.37140655517578, "decoder_belief_top1_on_support": 0.01953125, "mean_decoder_belief_support_size": 1.869140625, "mean_slot_purity": 0.7820848822593689, "mean_slot_entropy": 0.44480936229228973, "mean_weight_entropy": 1.540618658065796, "mean_route_entropy": 0.0003399010456632823, "mean_route_accuracy": 1.0, "mean_target_support_size": 1.869140625 }, { "model": "cyclic_transition_ladder", "variant": "pilot_primitive_router", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "primitive_router", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 500, "length": 12, "query_type": "B", "k": 6, "n": 512, "decoder_query_target_mass": 0.19815123826265335, "decoder_query_target_nll": 9.235969066619873, "decoder_query_top1_on_support": 0.19921875, "mean_decoder_query_support_size": 1.6875, "decoder_belief_target_mass": 0.04107248969376087, "decoder_belief_target_nll": 16.439472198486328, "decoder_belief_top1_on_support": 0.03515625, "mean_decoder_belief_support_size": 1.869140625, "mean_slot_purity": 0.6832375526428223, "mean_slot_entropy": 0.7783715426921844, "mean_weight_entropy": 1.030007004737854, "mean_route_entropy": 0.0003327196463942528, "mean_route_accuracy": 1.0, "mean_target_support_size": 1.869140625 }, { "model": "cyclic_transition_ladder", "variant": "pilot_primitive_router", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "primitive_router", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 500, "length": 12, "query_type": "B", "k": 9, "n": 512, "decoder_query_target_mass": 0.3446420133113861, "decoder_query_target_nll": 7.916715145111084, "decoder_query_top1_on_support": 0.341796875, "mean_decoder_query_support_size": 1.6875, "decoder_belief_target_mass": 0.09151436388492584, "decoder_belief_target_nll": 14.588720321655273, "decoder_belief_top1_on_support": 0.087890625, "mean_decoder_belief_support_size": 1.869140625, "mean_slot_purity": 0.6080763638019562, "mean_slot_entropy": 1.041651725769043, "mean_weight_entropy": 0.6776070594787598, "mean_route_entropy": 0.0003144221263937652, "mean_route_accuracy": 1.0, "mean_target_support_size": 1.869140625 }, { "model": "cyclic_transition_ladder", "variant": "pilot_primitive_router", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "primitive_router", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 500, "length": 12, "query_type": "B", "k": 12, "n": 512, "decoder_query_target_mass": 0.9998176097869873, "decoder_query_target_nll": 0.35756242275238037, "decoder_query_top1_on_support": 1.0, "mean_decoder_query_support_size": 1.6875, "decoder_belief_target_mass": 0.9996309280395508, "decoder_belief_target_nll": 0.43263714015483856, "decoder_belief_top1_on_support": 1.0, "mean_decoder_belief_support_size": 1.869140625, "mean_slot_purity": 0.5481719672679901, "mean_slot_entropy": 1.2698355913162231, "mean_weight_entropy": 0.4328983277082443, "mean_route_entropy": 0.0003645978285931051, "mean_route_accuracy": 1.0, "mean_target_support_size": 1.869140625 }, { "model": "cyclic_transition_ladder", "variant": "pilot_primitive_router", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "primitive_router", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 500, "length": 12, "query_type": "A_PLUS_B", "k": 0, "n": 512, "decoder_query_target_mass": 0.17116477340459824, "decoder_query_target_nll": 2.397895336151123, "decoder_query_top1_on_support": 0.158203125, "mean_decoder_query_support_size": 1.8828125, "decoder_belief_target_mass": 0.016512785106897354, "decoder_belief_target_nll": 16.462244987487793, "decoder_belief_top1_on_support": 0.017578125, "mean_decoder_belief_support_size": 2.046875, "mean_slot_purity": 0.999997615814209, "mean_slot_entropy": 3.839540295302868e-05, "mean_weight_entropy": 2.397895336151123, "mean_route_entropy": -1.0, "mean_route_accuracy": -1.0, "mean_target_support_size": 2.046875 }, { "model": "cyclic_transition_ladder", "variant": "pilot_primitive_router", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "primitive_router", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 500, "length": 12, "query_type": "A_PLUS_B", "k": 1, "n": 512, "decoder_query_target_mass": 0.16660723835229874, "decoder_query_target_nll": 5.411053657531738, "decoder_query_top1_on_support": 0.171875, "mean_decoder_query_support_size": 1.8828125, "decoder_belief_target_mass": 0.014796361327171326, "decoder_belief_target_nll": 16.71116065979004, "decoder_belief_top1_on_support": 0.013671875, "mean_decoder_belief_support_size": 2.046875, "mean_slot_purity": 0.8460983335971832, "mean_slot_entropy": 0.2438325583934784, "mean_weight_entropy": 1.92848539352417, "mean_route_entropy": 0.00034264056012034416, "mean_route_accuracy": 1.0, "mean_target_support_size": 2.046875 }, { "model": "cyclic_transition_ladder", "variant": "pilot_primitive_router", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "primitive_router", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 500, "length": 12, "query_type": "A_PLUS_B", "k": 2, "n": 512, "decoder_query_target_mass": 0.17140120267868042, "decoder_query_target_nll": 6.546042203903198, "decoder_query_top1_on_support": 0.1640625, "mean_decoder_query_support_size": 1.8828125, "decoder_belief_target_mass": 0.0176372192800045, "decoder_belief_target_nll": 16.370322227478027, "decoder_belief_top1_on_support": 0.009765625, "mean_decoder_belief_support_size": 2.046875, "mean_slot_purity": 0.7845884561538696, "mean_slot_entropy": 0.40126733481884, "mean_weight_entropy": 1.681140422821045, "mean_route_entropy": 0.0003381452552275732, "mean_route_accuracy": 1.0, "mean_target_support_size": 2.046875 }, { "model": "cyclic_transition_ladder", "variant": "pilot_primitive_router", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "primitive_router", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 500, "length": 12, "query_type": "A_PLUS_B", "k": 3, "n": 512, "decoder_query_target_mass": 0.17527761310338974, "decoder_query_target_nll": 7.505167484283447, "decoder_query_top1_on_support": 0.185546875, "mean_decoder_query_support_size": 1.8828125, "decoder_belief_target_mass": 0.024472585879266262, "decoder_belief_target_nll": 16.702856063842773, "decoder_belief_top1_on_support": 0.0234375, "mean_decoder_belief_support_size": 2.046875, "mean_slot_purity": 0.7430654764175415, "mean_slot_entropy": 0.5301997363567352, "mean_weight_entropy": 1.4525727033615112, "mean_route_entropy": 0.00032935975468717515, "mean_route_accuracy": 1.0, "mean_target_support_size": 2.046875 }, { "model": "cyclic_transition_ladder", "variant": "pilot_primitive_router", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "primitive_router", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 500, "length": 12, "query_type": "A_PLUS_B", "k": 6, "n": 512, "decoder_query_target_mass": 0.19649116694927216, "decoder_query_target_nll": 8.728946685791016, "decoder_query_top1_on_support": 0.203125, "mean_decoder_query_support_size": 1.8828125, "decoder_belief_target_mass": 0.036041634157299995, "decoder_belief_target_nll": 16.687639236450195, "decoder_belief_top1_on_support": 0.033203125, "mean_decoder_belief_support_size": 2.046875, "mean_slot_purity": 0.6635350584983826, "mean_slot_entropy": 0.8354028761386871, "mean_weight_entropy": 0.9927002787590027, "mean_route_entropy": 0.00031805800972506404, "mean_route_accuracy": 1.0, "mean_target_support_size": 2.046875 }, { "model": "cyclic_transition_ladder", "variant": "pilot_primitive_router", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "primitive_router", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 500, "length": 12, "query_type": "A_PLUS_B", "k": 9, "n": 512, "decoder_query_target_mass": 0.19456817209720612, "decoder_query_target_nll": 9.159387111663818, "decoder_query_top1_on_support": 0.208984375, "mean_decoder_query_support_size": 1.8828125, "decoder_belief_target_mass": 0.0804559625685215, "decoder_belief_target_nll": 14.950078964233398, "decoder_belief_top1_on_support": 0.083984375, "mean_decoder_belief_support_size": 2.046875, "mean_slot_purity": 0.599256694316864, "mean_slot_entropy": 1.083962857723236, "mean_weight_entropy": 0.67069411277771, "mean_route_entropy": 0.0003544995852280408, "mean_route_accuracy": 1.0, "mean_target_support_size": 2.046875 }, { "model": "cyclic_transition_ladder", "variant": "pilot_primitive_router", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "primitive_router", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 500, "length": 12, "query_type": "A_PLUS_B", "k": 12, "n": 512, "decoder_query_target_mass": 0.9996739327907562, "decoder_query_target_nll": 0.3962472975254059, "decoder_query_top1_on_support": 1.0, "mean_decoder_query_support_size": 1.8828125, "decoder_belief_target_mass": 0.9996057152748108, "decoder_belief_target_nll": 0.4595811367034912, "decoder_belief_top1_on_support": 1.0, "mean_decoder_belief_support_size": 2.046875, "mean_slot_purity": 0.5493814945220947, "mean_slot_entropy": 1.2705617547035217, "mean_weight_entropy": 0.45974430441856384, "mean_route_entropy": 0.0003338209498906508, "mean_route_accuracy": 1.0, "mean_target_support_size": 2.046875 }, { "model": "cyclic_transition_ladder", "variant": "pilot_primitive_router", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "primitive_router", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 500, "length": 12, "query_type": "A_MINUS_B", "k": 0, "n": 512, "decoder_query_target_mass": 0.16015625, "decoder_query_target_nll": 13.967867851257324, "decoder_query_top1_on_support": 0.16015625, "mean_decoder_query_support_size": 1.826171875, "decoder_belief_target_mass": 0.015447443351149559, "decoder_belief_target_nll": 16.365763664245605, "decoder_belief_top1_on_support": 0.01171875, "mean_decoder_belief_support_size": 1.923828125, "mean_slot_purity": 0.999997615814209, "mean_slot_entropy": 3.839540295302868e-05, "mean_weight_entropy": 2.397895336151123, "mean_route_entropy": -1.0, "mean_route_accuracy": -1.0, "mean_target_support_size": 1.923828125 }, { "model": "cyclic_transition_ladder", "variant": "pilot_primitive_router", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "primitive_router", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 500, "length": 12, "query_type": "A_MINUS_B", "k": 1, "n": 512, "decoder_query_target_mass": 0.17294061183929443, "decoder_query_target_nll": 9.228866577148438, "decoder_query_top1_on_support": 0.1640625, "mean_decoder_query_support_size": 1.826171875, "decoder_belief_target_mass": 0.017134208232164383, "decoder_belief_target_nll": 16.43437671661377, "decoder_belief_top1_on_support": 0.0234375, "mean_decoder_belief_support_size": 1.923828125, "mean_slot_purity": 0.861189216375351, "mean_slot_entropy": 0.21771842241287231, "mean_weight_entropy": 1.961798369884491, "mean_route_entropy": 0.0003095358988502994, "mean_route_accuracy": 1.0, "mean_target_support_size": 1.923828125 }, { "model": "cyclic_transition_ladder", "variant": "pilot_primitive_router", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "primitive_router", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 500, "length": 12, "query_type": "A_MINUS_B", "k": 2, "n": 512, "decoder_query_target_mass": 0.1592392921447754, "decoder_query_target_nll": 8.30039358139038, "decoder_query_top1_on_support": 0.1484375, "mean_decoder_query_support_size": 1.826171875, "decoder_belief_target_mass": 0.019087063148617744, "decoder_belief_target_nll": 16.41527271270752, "decoder_belief_top1_on_support": 0.0078125, "mean_decoder_belief_support_size": 1.923828125, "mean_slot_purity": 0.7911884188652039, "mean_slot_entropy": 0.3866989314556122, "mean_weight_entropy": 1.6690510511398315, "mean_route_entropy": 0.0003069808008149266, "mean_route_accuracy": 1.0, "mean_target_support_size": 1.923828125 }, { "model": "cyclic_transition_ladder", "variant": "pilot_primitive_router", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "primitive_router", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 500, "length": 12, "query_type": "A_MINUS_B", "k": 3, "n": 512, "decoder_query_target_mass": 0.17853204905986786, "decoder_query_target_nll": 8.337008476257324, "decoder_query_top1_on_support": 0.185546875, "mean_decoder_query_support_size": 1.826171875, "decoder_belief_target_mass": 0.01737064588814974, "decoder_belief_target_nll": 16.752065658569336, "decoder_belief_top1_on_support": 0.009765625, "mean_decoder_belief_support_size": 1.923828125, "mean_slot_purity": 0.7497270107269287, "mean_slot_entropy": 0.5099177956581116, "mean_weight_entropy": 1.4530644416809082, "mean_route_entropy": 0.0003367767931194976, "mean_route_accuracy": 1.0, "mean_target_support_size": 1.923828125 }, { "model": "cyclic_transition_ladder", "variant": "pilot_primitive_router", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "primitive_router", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 500, "length": 12, "query_type": "A_MINUS_B", "k": 6, "n": 512, "decoder_query_target_mass": 0.1646261066198349, "decoder_query_target_nll": 8.945409297943115, "decoder_query_top1_on_support": 0.162109375, "mean_decoder_query_support_size": 1.826171875, "decoder_belief_target_mass": 0.028616156429052353, "decoder_belief_target_nll": 16.359498023986816, "decoder_belief_top1_on_support": 0.033203125, "mean_decoder_belief_support_size": 1.923828125, "mean_slot_purity": 0.6669218838214874, "mean_slot_entropy": 0.8253877758979797, "mean_weight_entropy": 1.0262244641780853, "mean_route_entropy": 0.00035464990651234984, "mean_route_accuracy": 1.0, "mean_target_support_size": 1.923828125 }, { "model": "cyclic_transition_ladder", "variant": "pilot_primitive_router", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "primitive_router", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 500, "length": 12, "query_type": "A_MINUS_B", "k": 9, "n": 512, "decoder_query_target_mass": 0.19988922029733658, "decoder_query_target_nll": 8.884108066558838, "decoder_query_top1_on_support": 0.212890625, "mean_decoder_query_support_size": 1.826171875, "decoder_belief_target_mass": 0.09669359400868416, "decoder_belief_target_nll": 14.475086212158203, "decoder_belief_top1_on_support": 0.1015625, "mean_decoder_belief_support_size": 1.923828125, "mean_slot_purity": 0.59782874584198, "mean_slot_entropy": 1.0785201787948608, "mean_weight_entropy": 0.6994203925132751, "mean_route_entropy": 0.0003679155634017661, "mean_route_accuracy": 1.0, "mean_target_support_size": 1.923828125 }, { "model": "cyclic_transition_ladder", "variant": "pilot_primitive_router", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "primitive_router", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 500, "length": 12, "query_type": "A_MINUS_B", "k": 12, "n": 512, "decoder_query_target_mass": 0.9996720850467682, "decoder_query_target_nll": 0.3846765160560608, "decoder_query_top1_on_support": 1.0, "mean_decoder_query_support_size": 1.826171875, "decoder_belief_target_mass": 0.9996272921562195, "decoder_belief_target_nll": 0.4356905221939087, "decoder_belief_top1_on_support": 1.0, "mean_decoder_belief_support_size": 1.923828125, "mean_slot_purity": 0.544821560382843, "mean_slot_entropy": 1.2762298583984375, "mean_weight_entropy": 0.43591780960559845, "mean_route_entropy": 0.0003515441349009052, "mean_route_accuracy": 1.0, "mean_target_support_size": 1.923828125 } ] } ], "checkpoints": [ "large_artifacts/cyclic_transition_ladder/checkpoints/pilot_primitive_router/checkpoint_final.pt" ], "num_parameters": 35658 }