{ "args": { "variant_name": "pilot_mlp", "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": "mlp", "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": 900, "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": 150, "output_dir": "experiments/cyclic_transition_ladder/runs/pilot_mlp", "checkpoint_dir": "large_artifacts/cyclic_transition_ladder/checkpoints/pilot_mlp" }, "train": [ { "step": 1, "loss": 4.2664971351623535, "grad_norm": 0.0771026536822319, "elapsed_seconds": 0.698084831237793, "mean_slot_purity": 0.010546278208494186, "mean_weight_entropy": 2.397089719772339, "mean_route_entropy": -1.0, "mean_route_accuracy": -1.0, "belief_loss": 4.2664971351623535, "decoder_belief_target_mass": 0.04551476240158081, "decoder_belief_target_nll": 4.802532196044922, "decoder_belief_top1_on_support": 0.04296875, "mean_decoder_belief_support_size": 5.5234375, "decoder_query_target_mass": 0.4306308627128601, "decoder_query_target_nll": 2.39894962310791, "decoder_query_top1_on_support": 0.44140625, "mean_decoder_query_support_size": 4.734375 }, { "step": 150, "loss": 3.560971736907959, "grad_norm": 0.6879419088363647, "elapsed_seconds": 36.7713565826416, "mean_slot_purity": 0.07289133965969086, "mean_weight_entropy": 1.346082329750061, "mean_route_entropy": -1.0, "mean_route_accuracy": -1.0, "belief_loss": 3.560971736907959, "decoder_belief_target_mass": 0.13642162084579468, "decoder_belief_target_nll": 3.930468797683716, "decoder_belief_top1_on_support": 0.1953125, "mean_decoder_belief_support_size": 5.62890625, "decoder_query_target_mass": 0.5582504868507385, "decoder_query_target_nll": 2.0896525382995605, "decoder_query_top1_on_support": 0.60546875, "mean_decoder_query_support_size": 4.7734375 }, { "step": 300, "loss": 3.4237148761749268, "grad_norm": 0.7914518713951111, "elapsed_seconds": 74.07030177116394, "mean_slot_purity": 0.14058485627174377, "mean_weight_entropy": 1.3583102226257324, "mean_route_entropy": -1.0, "mean_route_accuracy": -1.0, "belief_loss": 3.4237148761749268, "decoder_belief_target_mass": 0.19900937378406525, "decoder_belief_target_nll": 3.744633674621582, "decoder_belief_top1_on_support": 0.2265625, "mean_decoder_belief_support_size": 5.359375, "decoder_query_target_mass": 0.5546859502792358, "decoder_query_target_nll": 2.045637369155884, "decoder_query_top1_on_support": 0.57421875, "mean_decoder_query_support_size": 4.5234375 }, { "step": 450, "loss": 3.2583279609680176, "grad_norm": 0.9258673787117004, "elapsed_seconds": 110.67611813545227, "mean_slot_purity": 0.15420886874198914, "mean_weight_entropy": 1.2839422225952148, "mean_route_entropy": -1.0, "mean_route_accuracy": -1.0, "belief_loss": 3.2583279609680176, "decoder_belief_target_mass": 0.222202867269516, "decoder_belief_target_nll": 3.648956775665283, "decoder_belief_top1_on_support": 0.30078125, "mean_decoder_belief_support_size": 6.2109375, "decoder_query_target_mass": 0.6281670331954956, "decoder_query_target_nll": 2.054920196533203, "decoder_query_top1_on_support": 0.66015625, "mean_decoder_query_support_size": 5.4765625 }, { "step": 600, "loss": 3.069988489151001, "grad_norm": 1.780191421508789, "elapsed_seconds": 147.0311255455017, "mean_slot_purity": 0.2299351841211319, "mean_weight_entropy": 1.1666442155838013, "mean_route_entropy": -1.0, "mean_route_accuracy": -1.0, "belief_loss": 3.069988489151001, "decoder_belief_target_mass": 0.2881419062614441, "decoder_belief_target_nll": 3.3733701705932617, "decoder_belief_top1_on_support": 0.40625, "mean_decoder_belief_support_size": 5.64453125, "decoder_query_target_mass": 0.6453896164894104, "decoder_query_target_nll": 1.8746196031570435, "decoder_query_top1_on_support": 0.6796875, "mean_decoder_query_support_size": 4.703125 }, { "step": 750, "loss": 3.0628843307495117, "grad_norm": 1.7043081521987915, "elapsed_seconds": 183.48681902885437, "mean_slot_purity": 0.24540697038173676, "mean_weight_entropy": 1.1723363399505615, "mean_route_entropy": -1.0, "mean_route_accuracy": -1.0, "belief_loss": 3.0628843307495117, "decoder_belief_target_mass": 0.30337250232696533, "decoder_belief_target_nll": 3.2845864295959473, "decoder_belief_top1_on_support": 0.453125, "mean_decoder_belief_support_size": 5.84375, "decoder_query_target_mass": 0.6546333432197571, "decoder_query_target_nll": 1.8523025512695312, "decoder_query_top1_on_support": 0.7890625, "mean_decoder_query_support_size": 4.87109375 }, { "step": 900, "loss": 2.7390966415405273, "grad_norm": 2.3059909343719482, "elapsed_seconds": 220.9977264404297, "mean_slot_purity": 0.3368886709213257, "mean_weight_entropy": 1.394332766532898, "mean_route_entropy": -1.0, "mean_route_accuracy": -1.0, "belief_loss": 2.7390966415405273, "decoder_belief_target_mass": 0.37584248185157776, "decoder_belief_target_nll": 2.935059070587158, "decoder_belief_top1_on_support": 0.671875, "mean_decoder_belief_support_size": 6.0859375, "decoder_query_target_mass": 0.7249504327774048, "decoder_query_target_nll": 1.7751418352127075, "decoder_query_top1_on_support": 0.84765625, "mean_decoder_query_support_size": 5.22265625 } ], "eval": [ { "step": 900, "rows": [ { "model": "cyclic_transition_ladder", "variant": "pilot_mlp", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "mlp", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 900, "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": 0.474609375, "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_mlp", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "mlp", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 900, "length": 3, "query_type": "A", "k": 1, "n": 512, "decoder_query_target_mass": 0.5710095167160034, "decoder_query_target_nll": 2.641139268875122, "decoder_query_top1_on_support": 0.58203125, "mean_decoder_query_support_size": 5.09765625, "decoder_belief_target_mass": 0.1250089444220066, "decoder_belief_target_nll": 6.6788249015808105, "decoder_belief_top1_on_support": 0.185546875, "mean_decoder_belief_support_size": 5.849609375, "mean_slot_purity": 0.4604683816432953, "mean_slot_entropy": 1.6709620952606201, "mean_weight_entropy": 1.794294536113739, "mean_route_entropy": -1.0, "mean_route_accuracy": -1.0, "mean_target_support_size": 5.849609375 }, { "model": "cyclic_transition_ladder", "variant": "pilot_mlp", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "mlp", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 900, "length": 3, "query_type": "A", "k": 2, "n": 512, "decoder_query_target_mass": 0.6566644310951233, "decoder_query_target_nll": 2.4681137800216675, "decoder_query_top1_on_support": 0.70703125, "mean_decoder_query_support_size": 5.09765625, "decoder_belief_target_mass": 0.16330303251743317, "decoder_belief_target_nll": 5.361906290054321, "decoder_belief_top1_on_support": 0.240234375, "mean_decoder_belief_support_size": 5.849609375, "mean_slot_purity": 0.34191039204597473, "mean_slot_entropy": 2.17477548122406, "mean_weight_entropy": 1.5487053990364075, "mean_route_entropy": -1.0, "mean_route_accuracy": -1.0, "mean_target_support_size": 5.849609375 }, { "model": "cyclic_transition_ladder", "variant": "pilot_mlp", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "mlp", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 900, "length": 3, "query_type": "A", "k": 3, "n": 512, "decoder_query_target_mass": 0.8005860149860382, "decoder_query_target_nll": 1.653353989124298, "decoder_query_top1_on_support": 0.92578125, "mean_decoder_query_support_size": 5.09765625, "decoder_belief_target_mass": 0.35097536444664, "decoder_belief_target_nll": 2.854838490486145, "decoder_belief_top1_on_support": 0.650390625, "mean_decoder_belief_support_size": 5.849609375, "mean_slot_purity": 0.31054410338401794, "mean_slot_entropy": 2.3380379676818848, "mean_weight_entropy": 1.3683876395225525, "mean_route_entropy": -1.0, "mean_route_accuracy": -1.0, "mean_target_support_size": 5.849609375 }, { "model": "cyclic_transition_ladder", "variant": "pilot_mlp", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "mlp", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 900, "length": 3, "query_type": "A", "k": 6, "n": 512, "decoder_query_target_mass": 0.8005860149860382, "decoder_query_target_nll": 1.653353989124298, "decoder_query_top1_on_support": 0.92578125, "mean_decoder_query_support_size": 5.09765625, "decoder_belief_target_mass": 0.35097536444664, "decoder_belief_target_nll": 2.854838490486145, "decoder_belief_top1_on_support": 0.650390625, "mean_decoder_belief_support_size": 5.849609375, "mean_slot_purity": 0.31054410338401794, "mean_slot_entropy": 2.3380379676818848, "mean_weight_entropy": 1.3683876395225525, "mean_route_entropy": -1.0, "mean_route_accuracy": -1.0, "mean_target_support_size": 5.849609375 }, { "model": "cyclic_transition_ladder", "variant": "pilot_mlp", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "mlp", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 900, "length": 3, "query_type": "A", "k": 9, "n": 512, "decoder_query_target_mass": 0.8005860149860382, "decoder_query_target_nll": 1.653353989124298, "decoder_query_top1_on_support": 0.92578125, "mean_decoder_query_support_size": 5.09765625, "decoder_belief_target_mass": 0.35097536444664, "decoder_belief_target_nll": 2.854838490486145, "decoder_belief_top1_on_support": 0.650390625, "mean_decoder_belief_support_size": 5.849609375, "mean_slot_purity": 0.31054410338401794, "mean_slot_entropy": 2.3380379676818848, "mean_weight_entropy": 1.3683876395225525, "mean_route_entropy": -1.0, "mean_route_accuracy": -1.0, "mean_target_support_size": 5.849609375 }, { "model": "cyclic_transition_ladder", "variant": "pilot_mlp", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "mlp", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 900, "length": 3, "query_type": "A", "k": 12, "n": 512, "decoder_query_target_mass": 0.8005860149860382, "decoder_query_target_nll": 1.653353989124298, "decoder_query_top1_on_support": 0.92578125, "mean_decoder_query_support_size": 5.09765625, "decoder_belief_target_mass": 0.35097536444664, "decoder_belief_target_nll": 2.854838490486145, "decoder_belief_top1_on_support": 0.650390625, "mean_decoder_belief_support_size": 5.849609375, "mean_slot_purity": 0.31054410338401794, "mean_slot_entropy": 2.3380379676818848, "mean_weight_entropy": 1.3683876395225525, "mean_route_entropy": -1.0, "mean_route_accuracy": -1.0, "mean_target_support_size": 5.849609375 }, { "model": "cyclic_transition_ladder", "variant": "pilot_mlp", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "mlp", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 900, "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": 0.466796875, "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_mlp", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "mlp", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 900, "length": 3, "query_type": "B", "k": 1, "n": 512, "decoder_query_target_mass": 0.5428090691566467, "decoder_query_target_nll": 2.626664638519287, "decoder_query_top1_on_support": 0.541015625, "mean_decoder_query_support_size": 5.005859375, "decoder_belief_target_mass": 0.11739984899759293, "decoder_belief_target_nll": 6.4098780155181885, "decoder_belief_top1_on_support": 0.14453125, "mean_decoder_belief_support_size": 5.912109375, "mean_slot_purity": 0.4387749582529068, "mean_slot_entropy": 1.7492985129356384, "mean_weight_entropy": 1.8276550769805908, "mean_route_entropy": -1.0, "mean_route_accuracy": -1.0, "mean_target_support_size": 5.912109375 }, { "model": "cyclic_transition_ladder", "variant": "pilot_mlp", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "mlp", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 900, "length": 3, "query_type": "B", "k": 2, "n": 512, "decoder_query_target_mass": 0.6519822478294373, "decoder_query_target_nll": 2.418755531311035, "decoder_query_top1_on_support": 0.6953125, "mean_decoder_query_support_size": 5.005859375, "decoder_belief_target_mass": 0.16176281869411469, "decoder_belief_target_nll": 5.530059337615967, "decoder_belief_top1_on_support": 0.2265625, "mean_decoder_belief_support_size": 5.912109375, "mean_slot_purity": 0.3501826971769333, "mean_slot_entropy": 2.1399365663528442, "mean_weight_entropy": 1.5455377697944641, "mean_route_entropy": -1.0, "mean_route_accuracy": -1.0, "mean_target_support_size": 5.912109375 }, { "model": "cyclic_transition_ladder", "variant": "pilot_mlp", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "mlp", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 900, "length": 3, "query_type": "B", "k": 3, "n": 512, "decoder_query_target_mass": 0.7968186736106873, "decoder_query_target_nll": 1.6078577637672424, "decoder_query_top1_on_support": 0.919921875, "mean_decoder_query_support_size": 5.005859375, "decoder_belief_target_mass": 0.3382623791694641, "decoder_belief_target_nll": 2.930523991584778, "decoder_belief_top1_on_support": 0.6640625, "mean_decoder_belief_support_size": 5.912109375, "mean_slot_purity": 0.3060499280691147, "mean_slot_entropy": 2.3481414318084717, "mean_weight_entropy": 1.3989315032958984, "mean_route_entropy": -1.0, "mean_route_accuracy": -1.0, "mean_target_support_size": 5.912109375 }, { "model": "cyclic_transition_ladder", "variant": "pilot_mlp", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "mlp", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 900, "length": 3, "query_type": "B", "k": 6, "n": 512, "decoder_query_target_mass": 0.7968186736106873, "decoder_query_target_nll": 1.6078578233718872, "decoder_query_top1_on_support": 0.919921875, "mean_decoder_query_support_size": 5.005859375, "decoder_belief_target_mass": 0.3382623791694641, "decoder_belief_target_nll": 2.930523991584778, "decoder_belief_top1_on_support": 0.6640625, "mean_decoder_belief_support_size": 5.912109375, "mean_slot_purity": 0.3060499280691147, "mean_slot_entropy": 2.3481414318084717, "mean_weight_entropy": 1.3989315032958984, "mean_route_entropy": -1.0, "mean_route_accuracy": -1.0, "mean_target_support_size": 5.912109375 }, { "model": "cyclic_transition_ladder", "variant": "pilot_mlp", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "mlp", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 900, "length": 3, "query_type": "B", "k": 9, "n": 512, "decoder_query_target_mass": 0.7968186438083649, "decoder_query_target_nll": 1.6078577637672424, "decoder_query_top1_on_support": 0.919921875, "mean_decoder_query_support_size": 5.005859375, "decoder_belief_target_mass": 0.3382623791694641, "decoder_belief_target_nll": 2.930523991584778, "decoder_belief_top1_on_support": 0.6640625, "mean_decoder_belief_support_size": 5.912109375, "mean_slot_purity": 0.3060499280691147, "mean_slot_entropy": 2.3481414318084717, "mean_weight_entropy": 1.3989315032958984, "mean_route_entropy": -1.0, "mean_route_accuracy": -1.0, "mean_target_support_size": 5.912109375 }, { "model": "cyclic_transition_ladder", "variant": "pilot_mlp", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "mlp", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 900, "length": 3, "query_type": "B", "k": 12, "n": 512, "decoder_query_target_mass": 0.7968186736106873, "decoder_query_target_nll": 1.6078577637672424, "decoder_query_top1_on_support": 0.919921875, "mean_decoder_query_support_size": 5.005859375, "decoder_belief_target_mass": 0.3382623791694641, "decoder_belief_target_nll": 2.930523991584778, "decoder_belief_top1_on_support": 0.6640625, "mean_decoder_belief_support_size": 5.912109375, "mean_slot_purity": 0.3060499280691147, "mean_slot_entropy": 2.3481414318084717, "mean_weight_entropy": 1.3989315032958984, "mean_route_entropy": -1.0, "mean_route_accuracy": -1.0, "mean_target_support_size": 5.912109375 }, { "model": "cyclic_transition_ladder", "variant": "pilot_mlp", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "mlp", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 900, "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": 0.51171875, "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_mlp", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "mlp", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 900, "length": 3, "query_type": "A_PLUS_B", "k": 1, "n": 512, "decoder_query_target_mass": 0.5567127466201782, "decoder_query_target_nll": 2.7422038316726685, "decoder_query_top1_on_support": 0.59375, "mean_decoder_query_support_size": 5.857421875, "decoder_belief_target_mass": 0.13377170264720917, "decoder_belief_target_nll": 6.488939523696899, "decoder_belief_top1_on_support": 0.1796875, "mean_decoder_belief_support_size": 6.25390625, "mean_slot_purity": 0.44133126735687256, "mean_slot_entropy": 1.74257093667984, "mean_weight_entropy": 1.8059601187705994, "mean_route_entropy": -1.0, "mean_route_accuracy": -1.0, "mean_target_support_size": 6.25390625 }, { "model": "cyclic_transition_ladder", "variant": "pilot_mlp", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "mlp", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 900, "length": 3, "query_type": "A_PLUS_B", "k": 2, "n": 512, "decoder_query_target_mass": 0.572236567735672, "decoder_query_target_nll": 2.6724928617477417, "decoder_query_top1_on_support": 0.59375, "mean_decoder_query_support_size": 5.857421875, "decoder_belief_target_mass": 0.1795182302594185, "decoder_belief_target_nll": 5.438790798187256, "decoder_belief_top1_on_support": 0.25390625, "mean_decoder_belief_support_size": 6.25390625, "mean_slot_purity": 0.34739725291728973, "mean_slot_entropy": 2.1579946279525757, "mean_weight_entropy": 1.557147204875946, "mean_route_entropy": -1.0, "mean_route_accuracy": -1.0, "mean_target_support_size": 6.25390625 }, { "model": "cyclic_transition_ladder", "variant": "pilot_mlp", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "mlp", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 900, "length": 3, "query_type": "A_PLUS_B", "k": 3, "n": 512, "decoder_query_target_mass": 0.6881932020187378, "decoder_query_target_nll": 2.035688877105713, "decoder_query_top1_on_support": 0.83984375, "mean_decoder_query_support_size": 5.857421875, "decoder_belief_target_mass": 0.36422140896320343, "decoder_belief_target_nll": 2.9361026287078857, "decoder_belief_top1_on_support": 0.64453125, "mean_decoder_belief_support_size": 6.25390625, "mean_slot_purity": 0.32071539759635925, "mean_slot_entropy": 2.275372266769409, "mean_weight_entropy": 1.3889235854148865, "mean_route_entropy": -1.0, "mean_route_accuracy": -1.0, "mean_target_support_size": 6.25390625 }, { "model": "cyclic_transition_ladder", "variant": "pilot_mlp", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "mlp", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 900, "length": 3, "query_type": "A_PLUS_B", "k": 6, "n": 512, "decoder_query_target_mass": 0.6881932020187378, "decoder_query_target_nll": 2.035688877105713, "decoder_query_top1_on_support": 0.83984375, "mean_decoder_query_support_size": 5.857421875, "decoder_belief_target_mass": 0.36422140896320343, "decoder_belief_target_nll": 2.9361026287078857, "decoder_belief_top1_on_support": 0.64453125, "mean_decoder_belief_support_size": 6.25390625, "mean_slot_purity": 0.32071539759635925, "mean_slot_entropy": 2.275372266769409, "mean_weight_entropy": 1.3889235854148865, "mean_route_entropy": -1.0, "mean_route_accuracy": -1.0, "mean_target_support_size": 6.25390625 }, { "model": "cyclic_transition_ladder", "variant": "pilot_mlp", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "mlp", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 900, "length": 3, "query_type": "A_PLUS_B", "k": 9, "n": 512, "decoder_query_target_mass": 0.6881932020187378, "decoder_query_target_nll": 2.035688877105713, "decoder_query_top1_on_support": 0.83984375, "mean_decoder_query_support_size": 5.857421875, "decoder_belief_target_mass": 0.36422140896320343, "decoder_belief_target_nll": 2.9361026287078857, "decoder_belief_top1_on_support": 0.64453125, "mean_decoder_belief_support_size": 6.25390625, "mean_slot_purity": 0.32071539759635925, "mean_slot_entropy": 2.275372266769409, "mean_weight_entropy": 1.3889235854148865, "mean_route_entropy": -1.0, "mean_route_accuracy": -1.0, "mean_target_support_size": 6.25390625 }, { "model": "cyclic_transition_ladder", "variant": "pilot_mlp", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "mlp", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 900, "length": 3, "query_type": "A_PLUS_B", "k": 12, "n": 512, "decoder_query_target_mass": 0.6881932020187378, "decoder_query_target_nll": 2.035688877105713, "decoder_query_top1_on_support": 0.83984375, "mean_decoder_query_support_size": 5.857421875, "decoder_belief_target_mass": 0.36422140896320343, "decoder_belief_target_nll": 2.9361026287078857, "decoder_belief_top1_on_support": 0.64453125, "mean_decoder_belief_support_size": 6.25390625, "mean_slot_purity": 0.32071539759635925, "mean_slot_entropy": 2.275372266769409, "mean_weight_entropy": 1.3889235854148865, "mean_route_entropy": -1.0, "mean_route_accuracy": -1.0, "mean_target_support_size": 6.25390625 }, { "model": "cyclic_transition_ladder", "variant": "pilot_mlp", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "mlp", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 900, "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_mlp", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "mlp", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 900, "length": 3, "query_type": "A_MINUS_B", "k": 1, "n": 512, "decoder_query_target_mass": 0.48058515787124634, "decoder_query_target_nll": 3.602400779724121, "decoder_query_top1_on_support": 0.509765625, "mean_decoder_query_support_size": 4.90625, "decoder_belief_target_mass": 0.14215736836194992, "decoder_belief_target_nll": 6.574612140655518, "decoder_belief_top1_on_support": 0.18359375, "mean_decoder_belief_support_size": 6.041015625, "mean_slot_purity": 0.44867900013923645, "mean_slot_entropy": 1.7288570404052734, "mean_weight_entropy": 1.760927438735962, "mean_route_entropy": -1.0, "mean_route_accuracy": -1.0, "mean_target_support_size": 6.041015625 }, { "model": "cyclic_transition_ladder", "variant": "pilot_mlp", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "mlp", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 900, "length": 3, "query_type": "A_MINUS_B", "k": 2, "n": 512, "decoder_query_target_mass": 0.5112932920455933, "decoder_query_target_nll": 2.93465256690979, "decoder_query_top1_on_support": 0.59375, "mean_decoder_query_support_size": 4.90625, "decoder_belief_target_mass": 0.19375235587358475, "decoder_belief_target_nll": 5.44014835357666, "decoder_belief_top1_on_support": 0.30078125, "mean_decoder_belief_support_size": 6.041015625, "mean_slot_purity": 0.35732924938201904, "mean_slot_entropy": 2.120170474052429, "mean_weight_entropy": 1.5019536018371582, "mean_route_entropy": -1.0, "mean_route_accuracy": -1.0, "mean_target_support_size": 6.041015625 }, { "model": "cyclic_transition_ladder", "variant": "pilot_mlp", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "mlp", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 900, "length": 3, "query_type": "A_MINUS_B", "k": 3, "n": 512, "decoder_query_target_mass": 0.6617326140403748, "decoder_query_target_nll": 1.764039695262909, "decoder_query_top1_on_support": 0.90234375, "mean_decoder_query_support_size": 4.90625, "decoder_belief_target_mass": 0.37546269595623016, "decoder_belief_target_nll": 2.881063222885132, "decoder_belief_top1_on_support": 0.720703125, "mean_decoder_belief_support_size": 6.041015625, "mean_slot_purity": 0.32344648241996765, "mean_slot_entropy": 2.283089280128479, "mean_weight_entropy": 1.3384329080581665, "mean_route_entropy": -1.0, "mean_route_accuracy": -1.0, "mean_target_support_size": 6.041015625 }, { "model": "cyclic_transition_ladder", "variant": "pilot_mlp", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "mlp", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 900, "length": 3, "query_type": "A_MINUS_B", "k": 6, "n": 512, "decoder_query_target_mass": 0.6617326140403748, "decoder_query_target_nll": 1.764039695262909, "decoder_query_top1_on_support": 0.90234375, "mean_decoder_query_support_size": 4.90625, "decoder_belief_target_mass": 0.37546269595623016, "decoder_belief_target_nll": 2.881063222885132, "decoder_belief_top1_on_support": 0.720703125, "mean_decoder_belief_support_size": 6.041015625, "mean_slot_purity": 0.32344648241996765, "mean_slot_entropy": 2.283089280128479, "mean_weight_entropy": 1.3384329080581665, "mean_route_entropy": -1.0, "mean_route_accuracy": -1.0, "mean_target_support_size": 6.041015625 }, { "model": "cyclic_transition_ladder", "variant": "pilot_mlp", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "mlp", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 900, "length": 3, "query_type": "A_MINUS_B", "k": 9, "n": 512, "decoder_query_target_mass": 0.6617326140403748, "decoder_query_target_nll": 1.764039695262909, "decoder_query_top1_on_support": 0.90234375, "mean_decoder_query_support_size": 4.90625, "decoder_belief_target_mass": 0.37546269595623016, "decoder_belief_target_nll": 2.881063222885132, "decoder_belief_top1_on_support": 0.720703125, "mean_decoder_belief_support_size": 6.041015625, "mean_slot_purity": 0.32344648241996765, "mean_slot_entropy": 2.283089280128479, "mean_weight_entropy": 1.3384329080581665, "mean_route_entropy": -1.0, "mean_route_accuracy": -1.0, "mean_target_support_size": 6.041015625 }, { "model": "cyclic_transition_ladder", "variant": "pilot_mlp", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "mlp", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 900, "length": 3, "query_type": "A_MINUS_B", "k": 12, "n": 512, "decoder_query_target_mass": 0.6617326140403748, "decoder_query_target_nll": 1.764039695262909, "decoder_query_top1_on_support": 0.90234375, "mean_decoder_query_support_size": 4.90625, "decoder_belief_target_mass": 0.37546269595623016, "decoder_belief_target_nll": 2.881063222885132, "decoder_belief_top1_on_support": 0.720703125, "mean_decoder_belief_support_size": 6.041015625, "mean_slot_purity": 0.32344648241996765, "mean_slot_entropy": 2.283089280128479, "mean_weight_entropy": 1.3384329080581665, "mean_route_entropy": -1.0, "mean_route_accuracy": -1.0, "mean_target_support_size": 6.041015625 }, { "model": "cyclic_transition_ladder", "variant": "pilot_mlp", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "mlp", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 900, "length": 6, "query_type": "A", "k": 0, "n": 512, "decoder_query_target_mass": 0.2736150473356247, "decoder_query_target_nll": 2.397895336151123, "decoder_query_top1_on_support": 0.267578125, "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_mlp", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "mlp", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 900, "length": 6, "query_type": "A", "k": 1, "n": 512, "decoder_query_target_mass": 0.2973228693008423, "decoder_query_target_nll": 3.130348801612854, "decoder_query_top1_on_support": 0.322265625, "mean_decoder_query_support_size": 3.009765625, "decoder_belief_target_mass": 0.03509826585650444, "decoder_belief_target_nll": 7.745211362838745, "decoder_belief_top1_on_support": 0.04296875, "mean_decoder_belief_support_size": 3.615234375, "mean_slot_purity": 0.4555877149105072, "mean_slot_entropy": 1.6983157396316528, "mean_weight_entropy": 1.7824885845184326, "mean_route_entropy": -1.0, "mean_route_accuracy": -1.0, "mean_target_support_size": 3.615234375 }, { "model": "cyclic_transition_ladder", "variant": "pilot_mlp", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "mlp", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 900, "length": 6, "query_type": "A", "k": 2, "n": 512, "decoder_query_target_mass": 0.31653885543346405, "decoder_query_target_nll": 3.639873743057251, "decoder_query_top1_on_support": 0.333984375, "mean_decoder_query_support_size": 3.009765625, "decoder_belief_target_mass": 0.04391562379896641, "decoder_belief_target_nll": 7.660702228546143, "decoder_belief_top1_on_support": 0.052734375, "mean_decoder_belief_support_size": 3.615234375, "mean_slot_purity": 0.3590630888938904, "mean_slot_entropy": 2.115786910057068, "mean_weight_entropy": 1.5216069221496582, "mean_route_entropy": -1.0, "mean_route_accuracy": -1.0, "mean_target_support_size": 3.615234375 }, { "model": "cyclic_transition_ladder", "variant": "pilot_mlp", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "mlp", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 900, "length": 6, "query_type": "A", "k": 3, "n": 512, "decoder_query_target_mass": 0.34774963557720184, "decoder_query_target_nll": 3.533933639526367, "decoder_query_top1_on_support": 0.37890625, "mean_decoder_query_support_size": 3.009765625, "decoder_belief_target_mass": 0.05369451642036438, "decoder_belief_target_nll": 7.381027460098267, "decoder_belief_top1_on_support": 0.08203125, "mean_decoder_belief_support_size": 3.615234375, "mean_slot_purity": 0.31032897531986237, "mean_slot_entropy": 2.333582639694214, "mean_weight_entropy": 1.384658396244049, "mean_route_entropy": -1.0, "mean_route_accuracy": -1.0, "mean_target_support_size": 3.615234375 }, { "model": "cyclic_transition_ladder", "variant": "pilot_mlp", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "mlp", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 900, "length": 6, "query_type": "A", "k": 6, "n": 512, "decoder_query_target_mass": 0.6124284565448761, "decoder_query_target_nll": 1.6064782738685608, "decoder_query_top1_on_support": 0.724609375, "mean_decoder_query_support_size": 3.009765625, "decoder_belief_target_mass": 0.2281988114118576, "decoder_belief_target_nll": 3.071855068206787, "decoder_belief_top1_on_support": 0.3828125, "mean_decoder_belief_support_size": 3.615234375, "mean_slot_purity": 0.28692999482154846, "mean_slot_entropy": 2.4478522539138794, "mean_weight_entropy": 1.0551102757453918, "mean_route_entropy": -1.0, "mean_route_accuracy": -1.0, "mean_target_support_size": 3.615234375 }, { "model": "cyclic_transition_ladder", "variant": "pilot_mlp", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "mlp", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 900, "length": 6, "query_type": "A", "k": 9, "n": 512, "decoder_query_target_mass": 0.6124284565448761, "decoder_query_target_nll": 1.6064782738685608, "decoder_query_top1_on_support": 0.724609375, "mean_decoder_query_support_size": 3.009765625, "decoder_belief_target_mass": 0.2281988114118576, "decoder_belief_target_nll": 3.071855068206787, "decoder_belief_top1_on_support": 0.3828125, "mean_decoder_belief_support_size": 3.615234375, "mean_slot_purity": 0.28692999482154846, "mean_slot_entropy": 2.4478522539138794, "mean_weight_entropy": 1.0551102757453918, "mean_route_entropy": -1.0, "mean_route_accuracy": -1.0, "mean_target_support_size": 3.615234375 }, { "model": "cyclic_transition_ladder", "variant": "pilot_mlp", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "mlp", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 900, "length": 6, "query_type": "A", "k": 12, "n": 512, "decoder_query_target_mass": 0.6124284565448761, "decoder_query_target_nll": 1.6064782738685608, "decoder_query_top1_on_support": 0.724609375, "mean_decoder_query_support_size": 3.009765625, "decoder_belief_target_mass": 0.2281988114118576, "decoder_belief_target_nll": 3.071855068206787, "decoder_belief_top1_on_support": 0.3828125, "mean_decoder_belief_support_size": 3.615234375, "mean_slot_purity": 0.28692999482154846, "mean_slot_entropy": 2.4478522539138794, "mean_weight_entropy": 1.0551102757453918, "mean_route_entropy": -1.0, "mean_route_accuracy": -1.0, "mean_target_support_size": 3.615234375 }, { "model": "cyclic_transition_ladder", "variant": "pilot_mlp", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "mlp", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 900, "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": 0.28125, "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_mlp", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "mlp", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 900, "length": 6, "query_type": "B", "k": 1, "n": 512, "decoder_query_target_mass": 0.3145744204521179, "decoder_query_target_nll": 3.113306999206543, "decoder_query_top1_on_support": 0.328125, "mean_decoder_query_support_size": 3.23046875, "decoder_belief_target_mass": 0.03832346014678478, "decoder_belief_target_nll": 7.696336269378662, "decoder_belief_top1_on_support": 0.044921875, "mean_decoder_belief_support_size": 3.689453125, "mean_slot_purity": 0.45311443507671356, "mean_slot_entropy": 1.6995450854301453, "mean_weight_entropy": 1.79181706905365, "mean_route_entropy": -1.0, "mean_route_accuracy": -1.0, "mean_target_support_size": 3.689453125 }, { "model": "cyclic_transition_ladder", "variant": "pilot_mlp", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "mlp", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 900, "length": 6, "query_type": "B", "k": 2, "n": 512, "decoder_query_target_mass": 0.34076160192489624, "decoder_query_target_nll": 3.516728162765503, "decoder_query_top1_on_support": 0.3671875, "mean_decoder_query_support_size": 3.23046875, "decoder_belief_target_mass": 0.04861043021082878, "decoder_belief_target_nll": 7.453635215759277, "decoder_belief_top1_on_support": 0.078125, "mean_decoder_belief_support_size": 3.689453125, "mean_slot_purity": 0.3519005626440048, "mean_slot_entropy": 2.1425530910491943, "mean_weight_entropy": 1.5775848627090454, "mean_route_entropy": -1.0, "mean_route_accuracy": -1.0, "mean_target_support_size": 3.689453125 }, { "model": "cyclic_transition_ladder", "variant": "pilot_mlp", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "mlp", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 900, "length": 6, "query_type": "B", "k": 3, "n": 512, "decoder_query_target_mass": 0.3593526929616928, "decoder_query_target_nll": 3.6144062280654907, "decoder_query_top1_on_support": 0.392578125, "mean_decoder_query_support_size": 3.23046875, "decoder_belief_target_mass": 0.05351003631949425, "decoder_belief_target_nll": 7.430673837661743, "decoder_belief_top1_on_support": 0.078125, "mean_decoder_belief_support_size": 3.689453125, "mean_slot_purity": 0.32486119866371155, "mean_slot_entropy": 2.286941885948181, "mean_weight_entropy": 1.394291639328003, "mean_route_entropy": -1.0, "mean_route_accuracy": -1.0, "mean_target_support_size": 3.689453125 }, { "model": "cyclic_transition_ladder", "variant": "pilot_mlp", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "mlp", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 900, "length": 6, "query_type": "B", "k": 6, "n": 512, "decoder_query_target_mass": 0.646836519241333, "decoder_query_target_nll": 1.6016210317611694, "decoder_query_top1_on_support": 0.7578125, "mean_decoder_query_support_size": 3.23046875, "decoder_belief_target_mass": 0.23745156824588776, "decoder_belief_target_nll": 3.116877555847168, "decoder_belief_top1_on_support": 0.392578125, "mean_decoder_belief_support_size": 3.689453125, "mean_slot_purity": 0.3105330765247345, "mean_slot_entropy": 2.3390400409698486, "mean_weight_entropy": 1.0499698519706726, "mean_route_entropy": -1.0, "mean_route_accuracy": -1.0, "mean_target_support_size": 3.689453125 }, { "model": "cyclic_transition_ladder", "variant": "pilot_mlp", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "mlp", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 900, "length": 6, "query_type": "B", "k": 9, "n": 512, "decoder_query_target_mass": 0.646836519241333, "decoder_query_target_nll": 1.6016210317611694, "decoder_query_top1_on_support": 0.7578125, "mean_decoder_query_support_size": 3.23046875, "decoder_belief_target_mass": 0.23745156824588776, "decoder_belief_target_nll": 3.116877555847168, "decoder_belief_top1_on_support": 0.392578125, "mean_decoder_belief_support_size": 3.689453125, "mean_slot_purity": 0.3105330765247345, "mean_slot_entropy": 2.3390400409698486, "mean_weight_entropy": 1.0499698519706726, "mean_route_entropy": -1.0, "mean_route_accuracy": -1.0, "mean_target_support_size": 3.689453125 }, { "model": "cyclic_transition_ladder", "variant": "pilot_mlp", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "mlp", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 900, "length": 6, "query_type": "B", "k": 12, "n": 512, "decoder_query_target_mass": 0.646836519241333, "decoder_query_target_nll": 1.6016210317611694, "decoder_query_top1_on_support": 0.7578125, "mean_decoder_query_support_size": 3.23046875, "decoder_belief_target_mass": 0.23745156824588776, "decoder_belief_target_nll": 3.116877555847168, "decoder_belief_top1_on_support": 0.392578125, "mean_decoder_belief_support_size": 3.689453125, "mean_slot_purity": 0.3105330765247345, "mean_slot_entropy": 2.3390400409698486, "mean_weight_entropy": 1.0499698519706726, "mean_route_entropy": -1.0, "mean_route_accuracy": -1.0, "mean_target_support_size": 3.689453125 }, { "model": "cyclic_transition_ladder", "variant": "pilot_mlp", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "mlp", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 900, "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": 0.294921875, "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_mlp", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "mlp", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 900, "length": 6, "query_type": "A_PLUS_B", "k": 1, "n": 512, "decoder_query_target_mass": 0.28771592676639557, "decoder_query_target_nll": 2.9409455060958862, "decoder_query_top1_on_support": 0.265625, "mean_decoder_query_support_size": 3.21875, "decoder_belief_target_mass": 0.04308287985622883, "decoder_belief_target_nll": 7.598062992095947, "decoder_belief_top1_on_support": 0.044921875, "mean_decoder_belief_support_size": 3.38671875, "mean_slot_purity": 0.46423766016960144, "mean_slot_entropy": 1.6741840839385986, "mean_weight_entropy": 1.7698071002960205, "mean_route_entropy": -1.0, "mean_route_accuracy": -1.0, "mean_target_support_size": 3.38671875 }, { "model": "cyclic_transition_ladder", "variant": "pilot_mlp", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "mlp", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 900, "length": 6, "query_type": "A_PLUS_B", "k": 2, "n": 512, "decoder_query_target_mass": 0.295355424284935, "decoder_query_target_nll": 2.949382185935974, "decoder_query_top1_on_support": 0.298828125, "mean_decoder_query_support_size": 3.21875, "decoder_belief_target_mass": 0.04935395531356335, "decoder_belief_target_nll": 7.570200204849243, "decoder_belief_top1_on_support": 0.064453125, "mean_decoder_belief_support_size": 3.38671875, "mean_slot_purity": 0.34858018159866333, "mean_slot_entropy": 2.1457602977752686, "mean_weight_entropy": 1.5073938369750977, "mean_route_entropy": -1.0, "mean_route_accuracy": -1.0, "mean_target_support_size": 3.38671875 }, { "model": "cyclic_transition_ladder", "variant": "pilot_mlp", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "mlp", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 900, "length": 6, "query_type": "A_PLUS_B", "k": 3, "n": 512, "decoder_query_target_mass": 0.29460906982421875, "decoder_query_target_nll": 3.0448381900787354, "decoder_query_top1_on_support": 0.29296875, "mean_decoder_query_support_size": 3.21875, "decoder_belief_target_mass": 0.06401215307414532, "decoder_belief_target_nll": 7.450846910476685, "decoder_belief_top1_on_support": 0.080078125, "mean_decoder_belief_support_size": 3.38671875, "mean_slot_purity": 0.32725030183792114, "mean_slot_entropy": 2.2635087966918945, "mean_weight_entropy": 1.3398829698562622, "mean_route_entropy": -1.0, "mean_route_accuracy": -1.0, "mean_target_support_size": 3.38671875 }, { "model": "cyclic_transition_ladder", "variant": "pilot_mlp", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "mlp", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 900, "length": 6, "query_type": "A_PLUS_B", "k": 6, "n": 512, "decoder_query_target_mass": 0.4558458626270294, "decoder_query_target_nll": 2.0077097415924072, "decoder_query_top1_on_support": 0.607421875, "mean_decoder_query_support_size": 3.21875, "decoder_belief_target_mass": 0.2507937103509903, "decoder_belief_target_nll": 3.0078072547912598, "decoder_belief_top1_on_support": 0.43359375, "mean_decoder_belief_support_size": 3.38671875, "mean_slot_purity": 0.3066447526216507, "mean_slot_entropy": 2.3576886653900146, "mean_weight_entropy": 1.004997968673706, "mean_route_entropy": -1.0, "mean_route_accuracy": -1.0, "mean_target_support_size": 3.38671875 }, { "model": "cyclic_transition_ladder", "variant": "pilot_mlp", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "mlp", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 900, "length": 6, "query_type": "A_PLUS_B", "k": 9, "n": 512, "decoder_query_target_mass": 0.4558458626270294, "decoder_query_target_nll": 2.0077097415924072, "decoder_query_top1_on_support": 0.607421875, "mean_decoder_query_support_size": 3.21875, "decoder_belief_target_mass": 0.2507937103509903, "decoder_belief_target_nll": 3.0078072547912598, "decoder_belief_top1_on_support": 0.43359375, "mean_decoder_belief_support_size": 3.38671875, "mean_slot_purity": 0.3066447526216507, "mean_slot_entropy": 2.3576886653900146, "mean_weight_entropy": 1.004997968673706, "mean_route_entropy": -1.0, "mean_route_accuracy": -1.0, "mean_target_support_size": 3.38671875 }, { "model": "cyclic_transition_ladder", "variant": "pilot_mlp", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "mlp", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 900, "length": 6, "query_type": "A_PLUS_B", "k": 12, "n": 512, "decoder_query_target_mass": 0.4558458626270294, "decoder_query_target_nll": 2.0077097415924072, "decoder_query_top1_on_support": 0.607421875, "mean_decoder_query_support_size": 3.21875, "decoder_belief_target_mass": 0.2507937103509903, "decoder_belief_target_nll": 3.0078072547912598, "decoder_belief_top1_on_support": 0.43359375, "mean_decoder_belief_support_size": 3.38671875, "mean_slot_purity": 0.3066447526216507, "mean_slot_entropy": 2.3576886653900146, "mean_weight_entropy": 1.004997968673706, "mean_route_entropy": -1.0, "mean_route_accuracy": -1.0, "mean_target_support_size": 3.38671875 }, { "model": "cyclic_transition_ladder", "variant": "pilot_mlp", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "mlp", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 900, "length": 6, "query_type": "A_MINUS_B", "k": 0, "n": 512, "decoder_query_target_mass": 0.3242186903953552, "decoder_query_target_nll": 13.719186782836914, "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_mlp", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "mlp", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 900, "length": 6, "query_type": "A_MINUS_B", "k": 1, "n": 512, "decoder_query_target_mass": 0.32292550802230835, "decoder_query_target_nll": 3.7321178913116455, "decoder_query_top1_on_support": 0.333984375, "mean_decoder_query_support_size": 3.408203125, "decoder_belief_target_mass": 0.04634442552924156, "decoder_belief_target_nll": 7.36961030960083, "decoder_belief_top1_on_support": 0.0390625, "mean_decoder_belief_support_size": 3.783203125, "mean_slot_purity": 0.45248597860336304, "mean_slot_entropy": 1.7074446082115173, "mean_weight_entropy": 1.7830212712287903, "mean_route_entropy": -1.0, "mean_route_accuracy": -1.0, "mean_target_support_size": 3.783203125 }, { "model": "cyclic_transition_ladder", "variant": "pilot_mlp", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "mlp", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 900, "length": 6, "query_type": "A_MINUS_B", "k": 2, "n": 512, "decoder_query_target_mass": 0.3065023422241211, "decoder_query_target_nll": 3.472049117088318, "decoder_query_top1_on_support": 0.31640625, "mean_decoder_query_support_size": 3.408203125, "decoder_belief_target_mass": 0.04960472136735916, "decoder_belief_target_nll": 7.578869342803955, "decoder_belief_top1_on_support": 0.05078125, "mean_decoder_belief_support_size": 3.783203125, "mean_slot_purity": 0.34967581927776337, "mean_slot_entropy": 2.1603342294692993, "mean_weight_entropy": 1.5105891227722168, "mean_route_entropy": -1.0, "mean_route_accuracy": -1.0, "mean_target_support_size": 3.783203125 }, { "model": "cyclic_transition_ladder", "variant": "pilot_mlp", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "mlp", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 900, "length": 6, "query_type": "A_MINUS_B", "k": 3, "n": 512, "decoder_query_target_mass": 0.32705725729465485, "decoder_query_target_nll": 3.129563331604004, "decoder_query_top1_on_support": 0.359375, "mean_decoder_query_support_size": 3.408203125, "decoder_belief_target_mass": 0.06220409832894802, "decoder_belief_target_nll": 7.396331787109375, "decoder_belief_top1_on_support": 0.080078125, "mean_decoder_belief_support_size": 3.783203125, "mean_slot_purity": 0.3186195194721222, "mean_slot_entropy": 2.3013076782226562, "mean_weight_entropy": 1.3491140604019165, "mean_route_entropy": -1.0, "mean_route_accuracy": -1.0, "mean_target_support_size": 3.783203125 }, { "model": "cyclic_transition_ladder", "variant": "pilot_mlp", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "mlp", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 900, "length": 6, "query_type": "A_MINUS_B", "k": 6, "n": 512, "decoder_query_target_mass": 0.4900405555963516, "decoder_query_target_nll": 1.8757294416427612, "decoder_query_top1_on_support": 0.62890625, "mean_decoder_query_support_size": 3.408203125, "decoder_belief_target_mass": 0.2534041255712509, "decoder_belief_target_nll": 3.0559449195861816, "decoder_belief_top1_on_support": 0.416015625, "mean_decoder_belief_support_size": 3.783203125, "mean_slot_purity": 0.2954772859811783, "mean_slot_entropy": 2.3965378999710083, "mean_weight_entropy": 1.0207568407058716, "mean_route_entropy": -1.0, "mean_route_accuracy": -1.0, "mean_target_support_size": 3.783203125 }, { "model": "cyclic_transition_ladder", "variant": "pilot_mlp", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "mlp", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 900, "length": 6, "query_type": "A_MINUS_B", "k": 9, "n": 512, "decoder_query_target_mass": 0.49004054069519043, "decoder_query_target_nll": 1.8757294416427612, "decoder_query_top1_on_support": 0.62890625, "mean_decoder_query_support_size": 3.408203125, "decoder_belief_target_mass": 0.2534041255712509, "decoder_belief_target_nll": 3.0559449195861816, "decoder_belief_top1_on_support": 0.416015625, "mean_decoder_belief_support_size": 3.783203125, "mean_slot_purity": 0.2954772859811783, "mean_slot_entropy": 2.3965378999710083, "mean_weight_entropy": 1.0207568407058716, "mean_route_entropy": -1.0, "mean_route_accuracy": -1.0, "mean_target_support_size": 3.783203125 }, { "model": "cyclic_transition_ladder", "variant": "pilot_mlp", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "mlp", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 900, "length": 6, "query_type": "A_MINUS_B", "k": 12, "n": 512, "decoder_query_target_mass": 0.4900405555963516, "decoder_query_target_nll": 1.8757294416427612, "decoder_query_top1_on_support": 0.62890625, "mean_decoder_query_support_size": 3.408203125, "decoder_belief_target_mass": 0.2534041255712509, "decoder_belief_target_nll": 3.0559449195861816, "decoder_belief_top1_on_support": 0.416015625, "mean_decoder_belief_support_size": 3.783203125, "mean_slot_purity": 0.2954772859811783, "mean_slot_entropy": 2.3965378999710083, "mean_weight_entropy": 1.0207568407058716, "mean_route_entropy": -1.0, "mean_route_accuracy": -1.0, "mean_target_support_size": 3.783203125 }, { "model": "cyclic_transition_ladder", "variant": "pilot_mlp", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "mlp", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 900, "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": 0.1953125, "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_mlp", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "mlp", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 900, "length": 9, "query_type": "A", "k": 1, "n": 512, "decoder_query_target_mass": 0.21425212919712067, "decoder_query_target_nll": 3.236292243003845, "decoder_query_top1_on_support": 0.220703125, "mean_decoder_query_support_size": 2.248046875, "decoder_belief_target_mass": 0.027524285949766636, "decoder_belief_target_nll": 7.656658172607422, "decoder_belief_top1_on_support": 0.033203125, "mean_decoder_belief_support_size": 2.623046875, "mean_slot_purity": 0.45098547637462616, "mean_slot_entropy": 1.7145289778709412, "mean_weight_entropy": 1.7652519941329956, "mean_route_entropy": -1.0, "mean_route_accuracy": -1.0, "mean_target_support_size": 2.623046875 }, { "model": "cyclic_transition_ladder", "variant": "pilot_mlp", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "mlp", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 900, "length": 9, "query_type": "A", "k": 2, "n": 512, "decoder_query_target_mass": 0.22159694880247116, "decoder_query_target_nll": 3.725531816482544, "decoder_query_top1_on_support": 0.21875, "mean_decoder_query_support_size": 2.248046875, "decoder_belief_target_mass": 0.02844698540866375, "decoder_belief_target_nll": 7.888422966003418, "decoder_belief_top1_on_support": 0.029296875, "mean_decoder_belief_support_size": 2.623046875, "mean_slot_purity": 0.3549259603023529, "mean_slot_entropy": 2.1175947189331055, "mean_weight_entropy": 1.5142700672149658, "mean_route_entropy": -1.0, "mean_route_accuracy": -1.0, "mean_target_support_size": 2.623046875 }, { "model": "cyclic_transition_ladder", "variant": "pilot_mlp", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "mlp", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 900, "length": 9, "query_type": "A", "k": 3, "n": 512, "decoder_query_target_mass": 0.2239832654595375, "decoder_query_target_nll": 4.161319971084595, "decoder_query_top1_on_support": 0.21484375, "mean_decoder_query_support_size": 2.248046875, "decoder_belief_target_mass": 0.03229381516575813, "decoder_belief_target_nll": 8.259620666503906, "decoder_belief_top1_on_support": 0.021484375, "mean_decoder_belief_support_size": 2.623046875, "mean_slot_purity": 0.3311486691236496, "mean_slot_entropy": 2.2503316402435303, "mean_weight_entropy": 1.322455644607544, "mean_route_entropy": -1.0, "mean_route_accuracy": -1.0, "mean_target_support_size": 2.623046875 }, { "model": "cyclic_transition_ladder", "variant": "pilot_mlp", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "mlp", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 900, "length": 9, "query_type": "A", "k": 6, "n": 512, "decoder_query_target_mass": 0.2796636372804642, "decoder_query_target_nll": 4.202930688858032, "decoder_query_top1_on_support": 0.314453125, "mean_decoder_query_support_size": 2.248046875, "decoder_belief_target_mass": 0.05476132407784462, "decoder_belief_target_nll": 8.21115255355835, "decoder_belief_top1_on_support": 0.08984375, "mean_decoder_belief_support_size": 2.623046875, "mean_slot_purity": 0.31472884118556976, "mean_slot_entropy": 2.322656512260437, "mean_weight_entropy": 1.0188348591327667, "mean_route_entropy": -1.0, "mean_route_accuracy": -1.0, "mean_target_support_size": 2.623046875 }, { "model": "cyclic_transition_ladder", "variant": "pilot_mlp", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "mlp", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 900, "length": 9, "query_type": "A", "k": 9, "n": 512, "decoder_query_target_mass": 0.5144985616207123, "decoder_query_target_nll": 1.6127336621284485, "decoder_query_top1_on_support": 0.625, "mean_decoder_query_support_size": 2.248046875, "decoder_belief_target_mass": 0.20590150356292725, "decoder_belief_target_nll": 3.142388701438904, "decoder_belief_top1_on_support": 0.3125, "mean_decoder_belief_support_size": 2.623046875, "mean_slot_purity": 0.30749230086803436, "mean_slot_entropy": 2.376380443572998, "mean_weight_entropy": 0.8211320042610168, "mean_route_entropy": -1.0, "mean_route_accuracy": -1.0, "mean_target_support_size": 2.623046875 }, { "model": "cyclic_transition_ladder", "variant": "pilot_mlp", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "mlp", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 900, "length": 9, "query_type": "A", "k": 12, "n": 512, "decoder_query_target_mass": 0.5144985616207123, "decoder_query_target_nll": 1.6127336621284485, "decoder_query_top1_on_support": 0.625, "mean_decoder_query_support_size": 2.248046875, "decoder_belief_target_mass": 0.20590150356292725, "decoder_belief_target_nll": 3.142388701438904, "decoder_belief_top1_on_support": 0.3125, "mean_decoder_belief_support_size": 2.623046875, "mean_slot_purity": 0.30749230086803436, "mean_slot_entropy": 2.376380443572998, "mean_weight_entropy": 0.8211320042610168, "mean_route_entropy": -1.0, "mean_route_accuracy": -1.0, "mean_target_support_size": 2.623046875 }, { "model": "cyclic_transition_ladder", "variant": "pilot_mlp", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "mlp", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 900, "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": 0.185546875, "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_mlp", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "mlp", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 900, "length": 9, "query_type": "B", "k": 1, "n": 512, "decoder_query_target_mass": 0.20414035767316818, "decoder_query_target_nll": 3.313072919845581, "decoder_query_top1_on_support": 0.21875, "mean_decoder_query_support_size": 2.138671875, "decoder_belief_target_mass": 0.021291740238666534, "decoder_belief_target_nll": 8.085602283477783, "decoder_belief_top1_on_support": 0.025390625, "mean_decoder_belief_support_size": 2.40234375, "mean_slot_purity": 0.46475332975387573, "mean_slot_entropy": 1.666229486465454, "mean_weight_entropy": 1.755608320236206, "mean_route_entropy": -1.0, "mean_route_accuracy": -1.0, "mean_target_support_size": 2.40234375 }, { "model": "cyclic_transition_ladder", "variant": "pilot_mlp", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "mlp", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 900, "length": 9, "query_type": "B", "k": 2, "n": 512, "decoder_query_target_mass": 0.21104910224676132, "decoder_query_target_nll": 3.9576025009155273, "decoder_query_top1_on_support": 0.205078125, "mean_decoder_query_support_size": 2.138671875, "decoder_belief_target_mass": 0.021175098605453968, "decoder_belief_target_nll": 8.173822402954102, "decoder_belief_top1_on_support": 0.015625, "mean_decoder_belief_support_size": 2.40234375, "mean_slot_purity": 0.36380015313625336, "mean_slot_entropy": 2.0828845500946045, "mean_weight_entropy": 1.4796051383018494, "mean_route_entropy": -1.0, "mean_route_accuracy": -1.0, "mean_target_support_size": 2.40234375 }, { "model": "cyclic_transition_ladder", "variant": "pilot_mlp", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "mlp", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 900, "length": 9, "query_type": "B", "k": 3, "n": 512, "decoder_query_target_mass": 0.20675040781497955, "decoder_query_target_nll": 4.26854133605957, "decoder_query_top1_on_support": 0.203125, "mean_decoder_query_support_size": 2.138671875, "decoder_belief_target_mass": 0.025050044059753418, "decoder_belief_target_nll": 8.32914400100708, "decoder_belief_top1_on_support": 0.025390625, "mean_decoder_belief_support_size": 2.40234375, "mean_slot_purity": 0.3236521780490875, "mean_slot_entropy": 2.2794713973999023, "mean_weight_entropy": 1.3077964782714844, "mean_route_entropy": -1.0, "mean_route_accuracy": -1.0, "mean_target_support_size": 2.40234375 }, { "model": "cyclic_transition_ladder", "variant": "pilot_mlp", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "mlp", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 900, "length": 9, "query_type": "B", "k": 6, "n": 512, "decoder_query_target_mass": 0.27342087030410767, "decoder_query_target_nll": 4.195162773132324, "decoder_query_top1_on_support": 0.294921875, "mean_decoder_query_support_size": 2.138671875, "decoder_belief_target_mass": 0.04526672512292862, "decoder_belief_target_nll": 8.081452369689941, "decoder_belief_top1_on_support": 0.0625, "mean_decoder_belief_support_size": 2.40234375, "mean_slot_purity": 0.2985345721244812, "mean_slot_entropy": 2.396110773086548, "mean_weight_entropy": 0.979778528213501, "mean_route_entropy": -1.0, "mean_route_accuracy": -1.0, "mean_target_support_size": 2.40234375 }, { "model": "cyclic_transition_ladder", "variant": "pilot_mlp", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "mlp", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 900, "length": 9, "query_type": "B", "k": 9, "n": 512, "decoder_query_target_mass": 0.5171160697937012, "decoder_query_target_nll": 1.592106819152832, "decoder_query_top1_on_support": 0.630859375, "mean_decoder_query_support_size": 2.138671875, "decoder_belief_target_mass": 0.19247162342071533, "decoder_belief_target_nll": 3.138371229171753, "decoder_belief_top1_on_support": 0.302734375, "mean_decoder_belief_support_size": 2.40234375, "mean_slot_purity": 0.3020141273736954, "mean_slot_entropy": 2.3746029138565063, "mean_weight_entropy": 0.7812629044055939, "mean_route_entropy": -1.0, "mean_route_accuracy": -1.0, "mean_target_support_size": 2.40234375 }, { "model": "cyclic_transition_ladder", "variant": "pilot_mlp", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "mlp", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 900, "length": 9, "query_type": "B", "k": 12, "n": 512, "decoder_query_target_mass": 0.5171160697937012, "decoder_query_target_nll": 1.5921068787574768, "decoder_query_top1_on_support": 0.630859375, "mean_decoder_query_support_size": 2.138671875, "decoder_belief_target_mass": 0.19247162342071533, "decoder_belief_target_nll": 3.138371229171753, "decoder_belief_top1_on_support": 0.302734375, "mean_decoder_belief_support_size": 2.40234375, "mean_slot_purity": 0.3020141273736954, "mean_slot_entropy": 2.3746029138565063, "mean_weight_entropy": 0.7812629044055939, "mean_route_entropy": -1.0, "mean_route_accuracy": -1.0, "mean_target_support_size": 2.40234375 }, { "model": "cyclic_transition_ladder", "variant": "pilot_mlp", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "mlp", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 900, "length": 9, "query_type": "A_PLUS_B", "k": 0, "n": 512, "decoder_query_target_mass": 0.22105822712183, "decoder_query_target_nll": 2.397895336151123, "decoder_query_top1_on_support": 0.236328125, "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_mlp", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "mlp", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 900, "length": 9, "query_type": "A_PLUS_B", "k": 1, "n": 512, "decoder_query_target_mass": 0.22458631545305252, "decoder_query_target_nll": 2.9365484714508057, "decoder_query_top1_on_support": 0.216796875, "mean_decoder_query_support_size": 2.431640625, "decoder_belief_target_mass": 0.0213249446824193, "decoder_belief_target_nll": 8.007402658462524, "decoder_belief_top1_on_support": 0.01953125, "mean_decoder_belief_support_size": 2.55078125, "mean_slot_purity": 0.4678906351327896, "mean_slot_entropy": 1.649766981601715, "mean_weight_entropy": 1.7870901823043823, "mean_route_entropy": -1.0, "mean_route_accuracy": -1.0, "mean_target_support_size": 2.55078125 }, { "model": "cyclic_transition_ladder", "variant": "pilot_mlp", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "mlp", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 900, "length": 9, "query_type": "A_PLUS_B", "k": 2, "n": 512, "decoder_query_target_mass": 0.22391197830438614, "decoder_query_target_nll": 2.9503320455551147, "decoder_query_top1_on_support": 0.19140625, "mean_decoder_query_support_size": 2.431640625, "decoder_belief_target_mass": 0.023619653657078743, "decoder_belief_target_nll": 7.967743635177612, "decoder_belief_top1_on_support": 0.017578125, "mean_decoder_belief_support_size": 2.55078125, "mean_slot_purity": 0.3575984984636307, "mean_slot_entropy": 2.1271175146102905, "mean_weight_entropy": 1.5342851281166077, "mean_route_entropy": -1.0, "mean_route_accuracy": -1.0, "mean_target_support_size": 2.55078125 }, { "model": "cyclic_transition_ladder", "variant": "pilot_mlp", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "mlp", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 900, "length": 9, "query_type": "A_PLUS_B", "k": 3, "n": 512, "decoder_query_target_mass": 0.2244163602590561, "decoder_query_target_nll": 2.9550284147262573, "decoder_query_top1_on_support": 0.212890625, "mean_decoder_query_support_size": 2.431640625, "decoder_belief_target_mass": 0.026573993265628815, "decoder_belief_target_nll": 8.328140258789062, "decoder_belief_top1_on_support": 0.03125, "mean_decoder_belief_support_size": 2.55078125, "mean_slot_purity": 0.3237955868244171, "mean_slot_entropy": 2.2866441011428833, "mean_weight_entropy": 1.3600786924362183, "mean_route_entropy": -1.0, "mean_route_accuracy": -1.0, "mean_target_support_size": 2.55078125 }, { "model": "cyclic_transition_ladder", "variant": "pilot_mlp", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "mlp", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 900, "length": 9, "query_type": "A_PLUS_B", "k": 6, "n": 512, "decoder_query_target_mass": 0.2152087241411209, "decoder_query_target_nll": 3.357210874557495, "decoder_query_top1_on_support": 0.205078125, "mean_decoder_query_support_size": 2.431640625, "decoder_belief_target_mass": 0.0431986078619957, "decoder_belief_target_nll": 8.120275974273682, "decoder_belief_top1_on_support": 0.0625, "mean_decoder_belief_support_size": 2.55078125, "mean_slot_purity": 0.3092516213655472, "mean_slot_entropy": 2.3403457403182983, "mean_weight_entropy": 1.0231585502624512, "mean_route_entropy": -1.0, "mean_route_accuracy": -1.0, "mean_target_support_size": 2.55078125 }, { "model": "cyclic_transition_ladder", "variant": "pilot_mlp", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "mlp", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 900, "length": 9, "query_type": "A_PLUS_B", "k": 9, "n": 512, "decoder_query_target_mass": 0.36771392822265625, "decoder_query_target_nll": 1.97865229845047, "decoder_query_top1_on_support": 0.478515625, "mean_decoder_query_support_size": 2.431640625, "decoder_belief_target_mass": 0.18037516623735428, "decoder_belief_target_nll": 3.2887613773345947, "decoder_belief_top1_on_support": 0.271484375, "mean_decoder_belief_support_size": 2.55078125, "mean_slot_purity": 0.29641929268836975, "mean_slot_entropy": 2.3839545249938965, "mean_weight_entropy": 0.8336102962493896, "mean_route_entropy": -1.0, "mean_route_accuracy": -1.0, "mean_target_support_size": 2.55078125 }, { "model": "cyclic_transition_ladder", "variant": "pilot_mlp", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "mlp", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 900, "length": 9, "query_type": "A_PLUS_B", "k": 12, "n": 512, "decoder_query_target_mass": 0.36771392822265625, "decoder_query_target_nll": 1.9786523580551147, "decoder_query_top1_on_support": 0.478515625, "mean_decoder_query_support_size": 2.431640625, "decoder_belief_target_mass": 0.18037516623735428, "decoder_belief_target_nll": 3.2887613773345947, "decoder_belief_top1_on_support": 0.271484375, "mean_decoder_belief_support_size": 2.55078125, "mean_slot_purity": 0.29641929268836975, "mean_slot_entropy": 2.3839545249938965, "mean_weight_entropy": 0.8336102962493896, "mean_route_entropy": -1.0, "mean_route_accuracy": -1.0, "mean_target_support_size": 2.55078125 }, { "model": "cyclic_transition_ladder", "variant": "pilot_mlp", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "mlp", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 900, "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_mlp", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "mlp", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 900, "length": 9, "query_type": "A_MINUS_B", "k": 1, "n": 512, "decoder_query_target_mass": 0.24618354439735413, "decoder_query_target_nll": 3.989401340484619, "decoder_query_top1_on_support": 0.236328125, "mean_decoder_query_support_size": 2.56640625, "decoder_belief_target_mass": 0.02790091373026371, "decoder_belief_target_nll": 7.991306304931641, "decoder_belief_top1_on_support": 0.029296875, "mean_decoder_belief_support_size": 2.8125, "mean_slot_purity": 0.45668990910053253, "mean_slot_entropy": 1.7018677592277527, "mean_weight_entropy": 1.7444801330566406, "mean_route_entropy": -1.0, "mean_route_accuracy": -1.0, "mean_target_support_size": 2.8125 }, { "model": "cyclic_transition_ladder", "variant": "pilot_mlp", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "mlp", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 900, "length": 9, "query_type": "A_MINUS_B", "k": 2, "n": 512, "decoder_query_target_mass": 0.23489795625209808, "decoder_query_target_nll": 3.4546886682510376, "decoder_query_top1_on_support": 0.232421875, "mean_decoder_query_support_size": 2.56640625, "decoder_belief_target_mass": 0.02725336141884327, "decoder_belief_target_nll": 8.057562828063965, "decoder_belief_top1_on_support": 0.029296875, "mean_decoder_belief_support_size": 2.8125, "mean_slot_purity": 0.337207555770874, "mean_slot_entropy": 2.2057546377182007, "mean_weight_entropy": 1.509311020374298, "mean_route_entropy": -1.0, "mean_route_accuracy": -1.0, "mean_target_support_size": 2.8125 }, { "model": "cyclic_transition_ladder", "variant": "pilot_mlp", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "mlp", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 900, "length": 9, "query_type": "A_MINUS_B", "k": 3, "n": 512, "decoder_query_target_mass": 0.23416344821453094, "decoder_query_target_nll": 3.216875672340393, "decoder_query_top1_on_support": 0.23046875, "mean_decoder_query_support_size": 2.56640625, "decoder_belief_target_mass": 0.02967068925499916, "decoder_belief_target_nll": 8.200425624847412, "decoder_belief_top1_on_support": 0.033203125, "mean_decoder_belief_support_size": 2.8125, "mean_slot_purity": 0.3008763939142227, "mean_slot_entropy": 2.360196590423584, "mean_weight_entropy": 1.3617608547210693, "mean_route_entropy": -1.0, "mean_route_accuracy": -1.0, "mean_target_support_size": 2.8125 }, { "model": "cyclic_transition_ladder", "variant": "pilot_mlp", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "mlp", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 900, "length": 9, "query_type": "A_MINUS_B", "k": 6, "n": 512, "decoder_query_target_mass": 0.23886432498693466, "decoder_query_target_nll": 3.2648080587387085, "decoder_query_top1_on_support": 0.232421875, "mean_decoder_query_support_size": 2.56640625, "decoder_belief_target_mass": 0.04885239526629448, "decoder_belief_target_nll": 7.8977556228637695, "decoder_belief_top1_on_support": 0.064453125, "mean_decoder_belief_support_size": 2.8125, "mean_slot_purity": 0.2984737455844879, "mean_slot_entropy": 2.399989604949951, "mean_weight_entropy": 1.041849136352539, "mean_route_entropy": -1.0, "mean_route_accuracy": -1.0, "mean_target_support_size": 2.8125 }, { "model": "cyclic_transition_ladder", "variant": "pilot_mlp", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "mlp", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 900, "length": 9, "query_type": "A_MINUS_B", "k": 9, "n": 512, "decoder_query_target_mass": 0.3945961892604828, "decoder_query_target_nll": 1.9401891231536865, "decoder_query_top1_on_support": 0.521484375, "mean_decoder_query_support_size": 2.56640625, "decoder_belief_target_mass": 0.18450210988521576, "decoder_belief_target_nll": 3.366178035736084, "decoder_belief_top1_on_support": 0.28515625, "mean_decoder_belief_support_size": 2.8125, "mean_slot_purity": 0.29939286410808563, "mean_slot_entropy": 2.4030139446258545, "mean_weight_entropy": 0.8646849393844604, "mean_route_entropy": -1.0, "mean_route_accuracy": -1.0, "mean_target_support_size": 2.8125 }, { "model": "cyclic_transition_ladder", "variant": "pilot_mlp", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "mlp", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 900, "length": 9, "query_type": "A_MINUS_B", "k": 12, "n": 512, "decoder_query_target_mass": 0.3945961892604828, "decoder_query_target_nll": 1.9401891231536865, "decoder_query_top1_on_support": 0.521484375, "mean_decoder_query_support_size": 2.56640625, "decoder_belief_target_mass": 0.18450210988521576, "decoder_belief_target_nll": 3.366178035736084, "decoder_belief_top1_on_support": 0.28515625, "mean_decoder_belief_support_size": 2.8125, "mean_slot_purity": 0.29939286410808563, "mean_slot_entropy": 2.4030139446258545, "mean_weight_entropy": 0.8646849393844604, "mean_route_entropy": -1.0, "mean_route_accuracy": -1.0, "mean_target_support_size": 2.8125 }, { "model": "cyclic_transition_ladder", "variant": "pilot_mlp", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "mlp", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 900, "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": 0.166015625, "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_mlp", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "mlp", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 900, "length": 12, "query_type": "A", "k": 1, "n": 512, "decoder_query_target_mass": 0.16318310797214508, "decoder_query_target_nll": 3.214214563369751, "decoder_query_top1_on_support": 0.16796875, "mean_decoder_query_support_size": 1.859375, "decoder_belief_target_mass": 0.013411436229944229, "decoder_belief_target_nll": 7.673417329788208, "decoder_belief_top1_on_support": 0.017578125, "mean_decoder_belief_support_size": 2.046875, "mean_slot_purity": 0.4477842450141907, "mean_slot_entropy": 1.7194081544876099, "mean_weight_entropy": 1.8372198939323425, "mean_route_entropy": -1.0, "mean_route_accuracy": -1.0, "mean_target_support_size": 2.046875 }, { "model": "cyclic_transition_ladder", "variant": "pilot_mlp", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "mlp", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 900, "length": 12, "query_type": "A", "k": 2, "n": 512, "decoder_query_target_mass": 0.1691589057445526, "decoder_query_target_nll": 3.7428956031799316, "decoder_query_top1_on_support": 0.17578125, "mean_decoder_query_support_size": 1.859375, "decoder_belief_target_mass": 0.013103279285132885, "decoder_belief_target_nll": 8.084481716156006, "decoder_belief_top1_on_support": 0.005859375, "mean_decoder_belief_support_size": 2.046875, "mean_slot_purity": 0.3511464595794678, "mean_slot_entropy": 2.1527148485183716, "mean_weight_entropy": 1.548355758190155, "mean_route_entropy": -1.0, "mean_route_accuracy": -1.0, "mean_target_support_size": 2.046875 }, { "model": "cyclic_transition_ladder", "variant": "pilot_mlp", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "mlp", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 900, "length": 12, "query_type": "A", "k": 3, "n": 512, "decoder_query_target_mass": 0.17476481199264526, "decoder_query_target_nll": 4.363329648971558, "decoder_query_top1_on_support": 0.185546875, "mean_decoder_query_support_size": 1.859375, "decoder_belief_target_mass": 0.015423694625496864, "decoder_belief_target_nll": 8.66731071472168, "decoder_belief_top1_on_support": 0.013671875, "mean_decoder_belief_support_size": 2.046875, "mean_slot_purity": 0.3258478045463562, "mean_slot_entropy": 2.258689284324646, "mean_weight_entropy": 1.3414602279663086, "mean_route_entropy": -1.0, "mean_route_accuracy": -1.0, "mean_target_support_size": 2.046875 }, { "model": "cyclic_transition_ladder", "variant": "pilot_mlp", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "mlp", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 900, "length": 12, "query_type": "A", "k": 6, "n": 512, "decoder_query_target_mass": 0.1946902498602867, "decoder_query_target_nll": 4.408613681793213, "decoder_query_top1_on_support": 0.19140625, "mean_decoder_query_support_size": 1.859375, "decoder_belief_target_mass": 0.02391548827290535, "decoder_belief_target_nll": 8.941696166992188, "decoder_belief_top1_on_support": 0.025390625, "mean_decoder_belief_support_size": 2.046875, "mean_slot_purity": 0.29878538846969604, "mean_slot_entropy": 2.384539008140564, "mean_weight_entropy": 1.0209410190582275, "mean_route_entropy": -1.0, "mean_route_accuracy": -1.0, "mean_target_support_size": 2.046875 }, { "model": "cyclic_transition_ladder", "variant": "pilot_mlp", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "mlp", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 900, "length": 12, "query_type": "A", "k": 9, "n": 512, "decoder_query_target_mass": 0.25267788767814636, "decoder_query_target_nll": 4.322619199752808, "decoder_query_top1_on_support": 0.28515625, "mean_decoder_query_support_size": 1.859375, "decoder_belief_target_mass": 0.03460695594549179, "decoder_belief_target_nll": 8.608136653900146, "decoder_belief_top1_on_support": 0.04296875, "mean_decoder_belief_support_size": 2.046875, "mean_slot_purity": 0.2995084673166275, "mean_slot_entropy": 2.376889705657959, "mean_weight_entropy": 0.7994803488254547, "mean_route_entropy": -1.0, "mean_route_accuracy": -1.0, "mean_target_support_size": 2.046875 }, { "model": "cyclic_transition_ladder", "variant": "pilot_mlp", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "mlp", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 900, "length": 12, "query_type": "A", "k": 12, "n": 512, "decoder_query_target_mass": 0.4515668749809265, "decoder_query_target_nll": 1.7706608772277832, "decoder_query_top1_on_support": 0.548828125, "mean_decoder_query_support_size": 1.859375, "decoder_belief_target_mass": 0.16039513051509857, "decoder_belief_target_nll": 3.4670095443725586, "decoder_belief_top1_on_support": 0.263671875, "mean_decoder_belief_support_size": 2.046875, "mean_slot_purity": 0.2998894155025482, "mean_slot_entropy": 2.3843871355056763, "mean_weight_entropy": 0.6889980137348175, "mean_route_entropy": -1.0, "mean_route_accuracy": -1.0, "mean_target_support_size": 2.046875 }, { "model": "cyclic_transition_ladder", "variant": "pilot_mlp", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "mlp", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 900, "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": 0.13671875, "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_mlp", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "mlp", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 900, "length": 12, "query_type": "B", "k": 1, "n": 512, "decoder_query_target_mass": 0.15312156826257706, "decoder_query_target_nll": 3.1694365739822388, "decoder_query_top1_on_support": 0.1328125, "mean_decoder_query_support_size": 1.6875, "decoder_belief_target_mass": 0.01709194015711546, "decoder_belief_target_nll": 7.473785877227783, "decoder_belief_top1_on_support": 0.013671875, "mean_decoder_belief_support_size": 1.869140625, "mean_slot_purity": 0.4342431128025055, "mean_slot_entropy": 1.766832947731018, "mean_weight_entropy": 1.802666187286377, "mean_route_entropy": -1.0, "mean_route_accuracy": -1.0, "mean_target_support_size": 1.869140625 }, { "model": "cyclic_transition_ladder", "variant": "pilot_mlp", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "mlp", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 900, "length": 12, "query_type": "B", "k": 2, "n": 512, "decoder_query_target_mass": 0.15002644062042236, "decoder_query_target_nll": 3.724911689758301, "decoder_query_top1_on_support": 0.1640625, "mean_decoder_query_support_size": 1.6875, "decoder_belief_target_mass": 0.017367888242006302, "decoder_belief_target_nll": 7.833638668060303, "decoder_belief_top1_on_support": 0.015625, "mean_decoder_belief_support_size": 1.869140625, "mean_slot_purity": 0.3350657671689987, "mean_slot_entropy": 2.207053303718567, "mean_weight_entropy": 1.5466286540031433, "mean_route_entropy": -1.0, "mean_route_accuracy": -1.0, "mean_target_support_size": 1.869140625 }, { "model": "cyclic_transition_ladder", "variant": "pilot_mlp", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "mlp", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 900, "length": 12, "query_type": "B", "k": 3, "n": 512, "decoder_query_target_mass": 0.15907685458660126, "decoder_query_target_nll": 4.195775508880615, "decoder_query_top1_on_support": 0.158203125, "mean_decoder_query_support_size": 1.6875, "decoder_belief_target_mass": 0.017107564955949783, "decoder_belief_target_nll": 8.365809917449951, "decoder_belief_top1_on_support": 0.017578125, "mean_decoder_belief_support_size": 1.869140625, "mean_slot_purity": 0.3108522742986679, "mean_slot_entropy": 2.3237192630767822, "mean_weight_entropy": 1.4122182726860046, "mean_route_entropy": -1.0, "mean_route_accuracy": -1.0, "mean_target_support_size": 1.869140625 }, { "model": "cyclic_transition_ladder", "variant": "pilot_mlp", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "mlp", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 900, "length": 12, "query_type": "B", "k": 6, "n": 512, "decoder_query_target_mass": 0.18195845931768417, "decoder_query_target_nll": 4.611596345901489, "decoder_query_top1_on_support": 0.193359375, "mean_decoder_query_support_size": 1.6875, "decoder_belief_target_mass": 0.02654606569558382, "decoder_belief_target_nll": 8.754618644714355, "decoder_belief_top1_on_support": 0.04296875, "mean_decoder_belief_support_size": 1.869140625, "mean_slot_purity": 0.3020395338535309, "mean_slot_entropy": 2.3779784440994263, "mean_weight_entropy": 1.074095070362091, "mean_route_entropy": -1.0, "mean_route_accuracy": -1.0, "mean_target_support_size": 1.869140625 }, { "model": "cyclic_transition_ladder", "variant": "pilot_mlp", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "mlp", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 900, "length": 12, "query_type": "B", "k": 9, "n": 512, "decoder_query_target_mass": 0.24029477685689926, "decoder_query_target_nll": 4.036052703857422, "decoder_query_top1_on_support": 0.2734375, "mean_decoder_query_support_size": 1.6875, "decoder_belief_target_mass": 0.04123810678720474, "decoder_belief_target_nll": 8.094935894012451, "decoder_belief_top1_on_support": 0.0546875, "mean_decoder_belief_support_size": 1.869140625, "mean_slot_purity": 0.3060702979564667, "mean_slot_entropy": 2.3521625995635986, "mean_weight_entropy": 0.8626508414745331, "mean_route_entropy": -1.0, "mean_route_accuracy": -1.0, "mean_target_support_size": 1.869140625 }, { "model": "cyclic_transition_ladder", "variant": "pilot_mlp", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "mlp", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 900, "length": 12, "query_type": "B", "k": 12, "n": 512, "decoder_query_target_mass": 0.4464917778968811, "decoder_query_target_nll": 1.7684813737869263, "decoder_query_top1_on_support": 0.53515625, "mean_decoder_query_support_size": 1.6875, "decoder_belief_target_mass": 0.16010832786560059, "decoder_belief_target_nll": 3.4810426235198975, "decoder_belief_top1_on_support": 0.228515625, "mean_decoder_belief_support_size": 1.869140625, "mean_slot_purity": 0.31206029653549194, "mean_slot_entropy": 2.3406238555908203, "mean_weight_entropy": 0.7181167900562286, "mean_route_entropy": -1.0, "mean_route_accuracy": -1.0, "mean_target_support_size": 1.869140625 }, { "model": "cyclic_transition_ladder", "variant": "pilot_mlp", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "mlp", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 900, "length": 12, "query_type": "A_PLUS_B", "k": 0, "n": 512, "decoder_query_target_mass": 0.17116478085517883, "decoder_query_target_nll": 2.397895336151123, "decoder_query_top1_on_support": 0.181640625, "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_mlp", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "mlp", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 900, "length": 12, "query_type": "A_PLUS_B", "k": 1, "n": 512, "decoder_query_target_mass": 0.1670682355761528, "decoder_query_target_nll": 2.986415982246399, "decoder_query_top1_on_support": 0.173828125, "mean_decoder_query_support_size": 1.8828125, "decoder_belief_target_mass": 0.015322274528443813, "decoder_belief_target_nll": 8.173203468322754, "decoder_belief_top1_on_support": 0.017578125, "mean_decoder_belief_support_size": 2.046875, "mean_slot_purity": 0.474543958902359, "mean_slot_entropy": 1.63417649269104, "mean_weight_entropy": 1.7604801058769226, "mean_route_entropy": -1.0, "mean_route_accuracy": -1.0, "mean_target_support_size": 2.046875 }, { "model": "cyclic_transition_ladder", "variant": "pilot_mlp", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "mlp", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 900, "length": 12, "query_type": "A_PLUS_B", "k": 2, "n": 512, "decoder_query_target_mass": 0.16420812159776688, "decoder_query_target_nll": 2.989962100982666, "decoder_query_top1_on_support": 0.16796875, "mean_decoder_query_support_size": 1.8828125, "decoder_belief_target_mass": 0.01730020996183157, "decoder_belief_target_nll": 8.308185577392578, "decoder_belief_top1_on_support": 0.017578125, "mean_decoder_belief_support_size": 2.046875, "mean_slot_purity": 0.356518417596817, "mean_slot_entropy": 2.135035514831543, "mean_weight_entropy": 1.509761929512024, "mean_route_entropy": -1.0, "mean_route_accuracy": -1.0, "mean_target_support_size": 2.046875 }, { "model": "cyclic_transition_ladder", "variant": "pilot_mlp", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "mlp", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 900, "length": 12, "query_type": "A_PLUS_B", "k": 3, "n": 512, "decoder_query_target_mass": 0.17047974467277527, "decoder_query_target_nll": 3.0510854721069336, "decoder_query_top1_on_support": 0.173828125, "mean_decoder_query_support_size": 1.8828125, "decoder_belief_target_mass": 0.018487350083887577, "decoder_belief_target_nll": 8.597072124481201, "decoder_belief_top1_on_support": 0.021484375, "mean_decoder_belief_support_size": 2.046875, "mean_slot_purity": 0.32541996240615845, "mean_slot_entropy": 2.2808313369750977, "mean_weight_entropy": 1.3345656394958496, "mean_route_entropy": -1.0, "mean_route_accuracy": -1.0, "mean_target_support_size": 2.046875 }, { "model": "cyclic_transition_ladder", "variant": "pilot_mlp", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "mlp", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 900, "length": 12, "query_type": "A_PLUS_B", "k": 6, "n": 512, "decoder_query_target_mass": 0.17295627295970917, "decoder_query_target_nll": 3.3481329679489136, "decoder_query_top1_on_support": 0.18359375, "mean_decoder_query_support_size": 1.8828125, "decoder_belief_target_mass": 0.023961187340319157, "decoder_belief_target_nll": 9.256710052490234, "decoder_belief_top1_on_support": 0.0234375, "mean_decoder_belief_support_size": 2.046875, "mean_slot_purity": 0.307381808757782, "mean_slot_entropy": 2.362618923187256, "mean_weight_entropy": 1.0053806900978088, "mean_route_entropy": -1.0, "mean_route_accuracy": -1.0, "mean_target_support_size": 2.046875 }, { "model": "cyclic_transition_ladder", "variant": "pilot_mlp", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "mlp", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 900, "length": 12, "query_type": "A_PLUS_B", "k": 9, "n": 512, "decoder_query_target_mass": 0.17158833891153336, "decoder_query_target_nll": 3.3761892318725586, "decoder_query_top1_on_support": 0.19140625, "mean_decoder_query_support_size": 1.8828125, "decoder_belief_target_mass": 0.037236252799630165, "decoder_belief_target_nll": 8.221019744873047, "decoder_belief_top1_on_support": 0.046875, "mean_decoder_belief_support_size": 2.046875, "mean_slot_purity": 0.28257495164871216, "mean_slot_entropy": 2.454477906227112, "mean_weight_entropy": 0.8200650811195374, "mean_route_entropy": -1.0, "mean_route_accuracy": -1.0, "mean_target_support_size": 2.046875 }, { "model": "cyclic_transition_ladder", "variant": "pilot_mlp", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "mlp", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 900, "length": 12, "query_type": "A_PLUS_B", "k": 12, "n": 512, "decoder_query_target_mass": 0.3187742382287979, "decoder_query_target_nll": 1.9806382060050964, "decoder_query_top1_on_support": 0.435546875, "mean_decoder_query_support_size": 1.8828125, "decoder_belief_target_mass": 0.17031338065862656, "decoder_belief_target_nll": 3.403204560279846, "decoder_belief_top1_on_support": 0.287109375, "mean_decoder_belief_support_size": 2.046875, "mean_slot_purity": 0.31141501665115356, "mean_slot_entropy": 2.3329488039016724, "mean_weight_entropy": 0.6907250285148621, "mean_route_entropy": -1.0, "mean_route_accuracy": -1.0, "mean_target_support_size": 2.046875 }, { "model": "cyclic_transition_ladder", "variant": "pilot_mlp", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "mlp", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 900, "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_mlp", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "mlp", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 900, "length": 12, "query_type": "A_MINUS_B", "k": 1, "n": 512, "decoder_query_target_mass": 0.1720120906829834, "decoder_query_target_nll": 3.883987545967102, "decoder_query_top1_on_support": 0.169921875, "mean_decoder_query_support_size": 1.826171875, "decoder_belief_target_mass": 0.01676116231828928, "decoder_belief_target_nll": 7.796436548233032, "decoder_belief_top1_on_support": 0.021484375, "mean_decoder_belief_support_size": 1.923828125, "mean_slot_purity": 0.4587889313697815, "mean_slot_entropy": 1.6910912990570068, "mean_weight_entropy": 1.746285617351532, "mean_route_entropy": -1.0, "mean_route_accuracy": -1.0, "mean_target_support_size": 1.923828125 }, { "model": "cyclic_transition_ladder", "variant": "pilot_mlp", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "mlp", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 900, "length": 12, "query_type": "A_MINUS_B", "k": 2, "n": 512, "decoder_query_target_mass": 0.16548651456832886, "decoder_query_target_nll": 3.500381350517273, "decoder_query_top1_on_support": 0.158203125, "mean_decoder_query_support_size": 1.826171875, "decoder_belief_target_mass": 0.018147466704249382, "decoder_belief_target_nll": 8.197741508483887, "decoder_belief_top1_on_support": 0.01953125, "mean_decoder_belief_support_size": 1.923828125, "mean_slot_purity": 0.35767772793769836, "mean_slot_entropy": 2.129371166229248, "mean_weight_entropy": 1.5010419487953186, "mean_route_entropy": -1.0, "mean_route_accuracy": -1.0, "mean_target_support_size": 1.923828125 }, { "model": "cyclic_transition_ladder", "variant": "pilot_mlp", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "mlp", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 900, "length": 12, "query_type": "A_MINUS_B", "k": 3, "n": 512, "decoder_query_target_mass": 0.1731514036655426, "decoder_query_target_nll": 3.338282823562622, "decoder_query_top1_on_support": 0.203125, "mean_decoder_query_support_size": 1.826171875, "decoder_belief_target_mass": 0.018704385496675968, "decoder_belief_target_nll": 8.454154968261719, "decoder_belief_top1_on_support": 0.017578125, "mean_decoder_belief_support_size": 1.923828125, "mean_slot_purity": 0.3152833729982376, "mean_slot_entropy": 2.3278485536575317, "mean_weight_entropy": 1.3301514983177185, "mean_route_entropy": -1.0, "mean_route_accuracy": -1.0, "mean_target_support_size": 1.923828125 }, { "model": "cyclic_transition_ladder", "variant": "pilot_mlp", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "mlp", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 900, "length": 12, "query_type": "A_MINUS_B", "k": 6, "n": 512, "decoder_query_target_mass": 0.16958708316087723, "decoder_query_target_nll": 3.3567516803741455, "decoder_query_top1_on_support": 0.16796875, "mean_decoder_query_support_size": 1.826171875, "decoder_belief_target_mass": 0.02277612965553999, "decoder_belief_target_nll": 8.88808012008667, "decoder_belief_top1_on_support": 0.021484375, "mean_decoder_belief_support_size": 1.923828125, "mean_slot_purity": 0.2965271770954132, "mean_slot_entropy": 2.4104251861572266, "mean_weight_entropy": 1.0218862295150757, "mean_route_entropy": -1.0, "mean_route_accuracy": -1.0, "mean_target_support_size": 1.923828125 }, { "model": "cyclic_transition_ladder", "variant": "pilot_mlp", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "mlp", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 900, "length": 12, "query_type": "A_MINUS_B", "k": 9, "n": 512, "decoder_query_target_mass": 0.16907337307929993, "decoder_query_target_nll": 3.518367290496826, "decoder_query_top1_on_support": 0.1796875, "mean_decoder_query_support_size": 1.826171875, "decoder_belief_target_mass": 0.03845973312854767, "decoder_belief_target_nll": 8.33298397064209, "decoder_belief_top1_on_support": 0.05859375, "mean_decoder_belief_support_size": 1.923828125, "mean_slot_purity": 0.30925513803958893, "mean_slot_entropy": 2.3536219596862793, "mean_weight_entropy": 0.8170657753944397, "mean_route_entropy": -1.0, "mean_route_accuracy": -1.0, "mean_target_support_size": 1.923828125 }, { "model": "cyclic_transition_ladder", "variant": "pilot_mlp", "modulus": 11, "observe_mod": 4, "observe_prob": 0.3, "op_family": "full", "query_family": "all", "slot_capacity": 11, "init_mode": "oracle", "transition_mode": "mlp", "supervision": "full_belief", "slot_dim": 64, "hidden_dim": 128, "fourier_order": 8, "train_steps": 900, "length": 12, "query_type": "A_MINUS_B", "k": 12, "n": 512, "decoder_query_target_mass": 0.31068263947963715, "decoder_query_target_nll": 1.9836099743843079, "decoder_query_top1_on_support": 0.419921875, "mean_decoder_query_support_size": 1.826171875, "decoder_belief_target_mass": 0.16337215900421143, "decoder_belief_target_nll": 3.3985488414764404, "decoder_belief_top1_on_support": 0.25, "mean_decoder_belief_support_size": 1.923828125, "mean_slot_purity": 0.31236863136291504, "mean_slot_entropy": 2.3214800357818604, "mean_weight_entropy": 0.6761638522148132, "mean_route_entropy": -1.0, "mean_route_accuracy": -1.0, "mean_target_support_size": 1.923828125 } ] } ], "checkpoints": [ "large_artifacts/cyclic_transition_ladder/checkpoints/pilot_mlp/checkpoint_final.pt" ], "num_parameters": 55127 }