{
  "args": {
    "variant_name": "smoke_cyclic_mixer",
    "seed": 83,
    "device": "auto",
    "modulus": 7,
    "observe_mod": 3,
    "observe_prob": 0.4,
    "op_family": "full",
    "query_family": "all",
    "slot_capacity": 7,
    "init_mode": "oracle",
    "transition_mode": "cyclic_mixer",
    "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": 3,
    "eval_lengths": "2,3",
    "eval_k": "0,1,2,3",
    "eval_query_types": "all",
    "train_steps": 500,
    "batch_size": 128,
    "eval_batch_size": 128,
    "eval_examples": 128,
    "lr": 0.001,
    "weight_decay": 0.0001,
    "belief_loss_weight": 1.0,
    "query_loss_weight": 0.0,
    "slot_entropy_weight": 0.0,
    "max_grad_norm": 1.0,
    "log_every": 100,
    "output_dir": "experiments/cyclic_transition_ladder/runs/smoke_cyclic_mixer",
    "checkpoint_dir": "large_artifacts/cyclic_transition_ladder/checkpoints/smoke_cyclic_mixer"
  },
  "train": [
    {
      "step": 1,
      "loss": 3.42120099067688,
      "grad_norm": 0.38171976804733276,
      "elapsed_seconds": 0.5108451843261719,
      "mean_slot_purity": 0.13619902729988098,
      "mean_weight_entropy": 1.0389317274093628,
      "mean_route_entropy": 1.5591609477996826,
      "mean_route_accuracy": 0.1380208432674408,
      "belief_loss": 3.42120099067688,
      "decoder_belief_target_mass": 0.1816735714673996,
      "decoder_belief_target_nll": 4.243255615234375,
      "decoder_belief_top1_on_support": 0.328125,
      "mean_decoder_belief_support_size": 4.3984375,
      "decoder_query_target_mass": 0.5829183459281921,
      "decoder_query_target_nll": 2.316802501678467,
      "decoder_query_top1_on_support": 0.640625,
      "mean_decoder_query_support_size": 3.8359375
    },
    {
      "step": 100,
      "loss": 1.544953465461731,
      "grad_norm": 0.0037990245036780834,
      "elapsed_seconds": 8.53774118423462,
      "mean_slot_purity": 0.820388913154602,
      "mean_weight_entropy": 1.2703626155853271,
      "mean_route_entropy": 0.06375432759523392,
      "mean_route_accuracy": 0.9166666865348816,
      "belief_loss": 1.544953465461731,
      "decoder_belief_target_mass": 0.9991468191146851,
      "decoder_belief_target_nll": 1.270571231842041,
      "decoder_belief_top1_on_support": 1.0,
      "mean_decoder_belief_support_size": 4.25,
      "decoder_query_target_mass": 0.9995510578155518,
      "decoder_query_target_nll": 0.9251920580863953,
      "decoder_query_top1_on_support": 1.0,
      "mean_decoder_query_support_size": 3.296875
    },
    {
      "step": 200,
      "loss": 1.5984183549880981,
      "grad_norm": 0.0016964853275567293,
      "elapsed_seconds": 16.368721961975098,
      "mean_slot_purity": 0.8370107412338257,
      "mean_weight_entropy": 1.3729119300842285,
      "mean_route_entropy": 0.0478837825357914,
      "mean_route_accuracy": 0.8932291865348816,
      "belief_loss": 1.5984183549880981,
      "decoder_belief_target_mass": 0.9995263814926147,
      "decoder_belief_target_nll": 1.373173475265503,
      "decoder_belief_top1_on_support": 1.0,
      "mean_decoder_belief_support_size": 4.6328125,
      "decoder_query_target_mass": 0.9998657703399658,
      "decoder_query_target_nll": 1.1627280712127686,
      "decoder_query_top1_on_support": 1.0,
      "mean_decoder_query_support_size": 4.1171875
    },
    {
      "step": 300,
      "loss": 1.5572320222854614,
      "grad_norm": 0.0008819213835522532,
      "elapsed_seconds": 23.99331545829773,
      "mean_slot_purity": 0.8283877372741699,
      "mean_weight_entropy": 1.2995980978012085,
      "mean_route_entropy": 0.0447327122092247,
      "mean_route_accuracy": 0.9166666865348816,
      "belief_loss": 1.5572320222854614,
      "decoder_belief_target_mass": 0.9997568130493164,
      "decoder_belief_target_nll": 1.2995789051055908,
      "decoder_belief_top1_on_support": 1.0,
      "mean_decoder_belief_support_size": 4.3828125,
      "decoder_query_target_mass": 0.999899685382843,
      "decoder_query_target_nll": 1.0458488464355469,
      "decoder_query_top1_on_support": 1.0,
      "mean_decoder_query_support_size": 3.7578125
    },
    {
      "step": 400,
      "loss": 1.5824588537216187,
      "grad_norm": 0.0006798719405196607,
      "elapsed_seconds": 31.5622820854187,
      "mean_slot_purity": 0.8308010101318359,
      "mean_weight_entropy": 1.3014260530471802,
      "mean_route_entropy": 0.047421813011169434,
      "mean_route_accuracy": 0.9140625,
      "belief_loss": 1.5824588537216187,
      "decoder_belief_target_mass": 0.9998279809951782,
      "decoder_belief_target_nll": 1.3014711141586304,
      "decoder_belief_top1_on_support": 1.0,
      "mean_decoder_belief_support_size": 4.3671875,
      "decoder_query_target_mass": 0.9999343156814575,
      "decoder_query_target_nll": 1.0783048868179321,
      "decoder_query_top1_on_support": 1.0,
      "mean_decoder_query_support_size": 3.8046875
    },
    {
      "step": 500,
      "loss": 1.5473524332046509,
      "grad_norm": 0.0004360081220511347,
      "elapsed_seconds": 39.13285946846008,
      "mean_slot_purity": 0.8006670475006104,
      "mean_weight_entropy": 1.275207281112671,
      "mean_route_entropy": 0.03958667814731598,
      "mean_route_accuracy": 0.9036458730697632,
      "belief_loss": 1.5473524332046509,
      "decoder_belief_target_mass": 0.9998850226402283,
      "decoder_belief_target_nll": 1.2752554416656494,
      "decoder_belief_top1_on_support": 1.0,
      "mean_decoder_belief_support_size": 4.3359375,
      "decoder_query_target_mass": 0.9999545812606812,
      "decoder_query_target_nll": 0.9920764565467834,
      "decoder_query_top1_on_support": 1.0,
      "mean_decoder_query_support_size": 3.546875
    }
  ],
  "eval": [
    {
      "step": 500,
      "rows": [
        {
          "model": "cyclic_transition_ladder",
          "variant": "smoke_cyclic_mixer",
          "modulus": 7,
          "observe_mod": 3,
          "observe_prob": 0.4,
          "op_family": "full",
          "query_family": "all",
          "slot_capacity": 7,
          "init_mode": "oracle",
          "transition_mode": "cyclic_mixer",
          "supervision": "full_belief",
          "slot_dim": 64,
          "hidden_dim": 128,
          "fourier_order": 8,
          "train_steps": 500,
          "length": 2,
          "query_type": "A",
          "k": 0,
          "n": 128,
          "decoder_query_target_mass": 0.495535671710968,
          "decoder_query_target_nll": 1.9459102153778076,
          "decoder_query_top1_on_support": 0.515625,
          "mean_decoder_query_support_size": 3.46875,
          "decoder_belief_target_mass": 0.13281241059303284,
          "decoder_belief_target_nll": 13.078685760498047,
          "decoder_belief_top1_on_support": 0.1875,
          "mean_decoder_belief_support_size": 4.1796875,
          "mean_slot_purity": 0.9999986290931702,
          "mean_slot_entropy": 2.3037249775370583e-05,
          "mean_weight_entropy": 1.9459102153778076,
          "mean_route_entropy": -1.0,
          "mean_route_accuracy": -1.0,
          "mean_target_support_size": 4.1796875
        },
        {
          "model": "cyclic_transition_ladder",
          "variant": "smoke_cyclic_mixer",
          "modulus": 7,
          "observe_mod": 3,
          "observe_prob": 0.4,
          "op_family": "full",
          "query_family": "all",
          "slot_capacity": 7,
          "init_mode": "oracle",
          "transition_mode": "cyclic_mixer",
          "supervision": "full_belief",
          "slot_dim": 64,
          "hidden_dim": 128,
          "fourier_order": 8,
          "train_steps": 500,
          "length": 2,
          "query_type": "A",
          "k": 1,
          "n": 128,
          "decoder_query_target_mass": 0.7094178199768066,
          "decoder_query_target_nll": 2.6614418029785156,
          "decoder_query_top1_on_support": 0.7265625,
          "mean_decoder_query_support_size": 3.46875,
          "decoder_belief_target_mass": 0.3424113988876343,
          "decoder_belief_target_nll": 8.200716972351074,
          "decoder_belief_top1_on_support": 0.3984375,
          "mean_decoder_belief_support_size": 4.1796875,
          "mean_slot_purity": 0.8596605658531189,
          "mean_slot_entropy": 0.21047429740428925,
          "mean_weight_entropy": 1.5265254974365234,
          "mean_route_entropy": 0.037843257188797,
          "mean_route_accuracy": 0.9010416865348816,
          "mean_target_support_size": 4.1796875
        },
        {
          "model": "cyclic_transition_ladder",
          "variant": "smoke_cyclic_mixer",
          "modulus": 7,
          "observe_mod": 3,
          "observe_prob": 0.4,
          "op_family": "full",
          "query_family": "all",
          "slot_capacity": 7,
          "init_mode": "oracle",
          "transition_mode": "cyclic_mixer",
          "supervision": "full_belief",
          "slot_dim": 64,
          "hidden_dim": 128,
          "fourier_order": 8,
          "train_steps": 500,
          "length": 2,
          "query_type": "A",
          "k": 2,
          "n": 128,
          "decoder_query_target_mass": 0.9999605417251587,
          "decoder_query_target_nll": 1.0153727531433105,
          "decoder_query_top1_on_support": 1.0,
          "mean_decoder_query_support_size": 3.46875,
          "decoder_belief_target_mass": 0.9998695254325867,
          "decoder_belief_target_nll": 1.2617096900939941,
          "decoder_belief_top1_on_support": 1.0,
          "mean_decoder_belief_support_size": 4.1796875,
          "mean_slot_purity": 0.8030068874359131,
          "mean_slot_entropy": 0.351245641708374,
          "mean_weight_entropy": 1.261706829071045,
          "mean_route_entropy": 0.035683631896972656,
          "mean_route_accuracy": 0.8984375,
          "mean_target_support_size": 4.1796875
        },
        {
          "model": "cyclic_transition_ladder",
          "variant": "smoke_cyclic_mixer",
          "modulus": 7,
          "observe_mod": 3,
          "observe_prob": 0.4,
          "op_family": "full",
          "query_family": "all",
          "slot_capacity": 7,
          "init_mode": "oracle",
          "transition_mode": "cyclic_mixer",
          "supervision": "full_belief",
          "slot_dim": 64,
          "hidden_dim": 128,
          "fourier_order": 8,
          "train_steps": 500,
          "length": 2,
          "query_type": "A",
          "k": 3,
          "n": 128,
          "decoder_query_target_mass": 0.9999605417251587,
          "decoder_query_target_nll": 1.0153727531433105,
          "decoder_query_top1_on_support": 1.0,
          "mean_decoder_query_support_size": 3.46875,
          "decoder_belief_target_mass": 0.9998695254325867,
          "decoder_belief_target_nll": 1.2617096900939941,
          "decoder_belief_top1_on_support": 1.0,
          "mean_decoder_belief_support_size": 4.1796875,
          "mean_slot_purity": 0.8030068874359131,
          "mean_slot_entropy": 0.351245641708374,
          "mean_weight_entropy": 1.261706829071045,
          "mean_route_entropy": -1.0,
          "mean_route_accuracy": -1.0,
          "mean_target_support_size": 4.1796875
        },
        {
          "model": "cyclic_transition_ladder",
          "variant": "smoke_cyclic_mixer",
          "modulus": 7,
          "observe_mod": 3,
          "observe_prob": 0.4,
          "op_family": "full",
          "query_family": "all",
          "slot_capacity": 7,
          "init_mode": "oracle",
          "transition_mode": "cyclic_mixer",
          "supervision": "full_belief",
          "slot_dim": 64,
          "hidden_dim": 128,
          "fourier_order": 8,
          "train_steps": 500,
          "length": 2,
          "query_type": "B",
          "k": 0,
          "n": 128,
          "decoder_query_target_mass": 0.5535714030265808,
          "decoder_query_target_nll": 1.9459102153778076,
          "decoder_query_top1_on_support": 0.5234375,
          "mean_decoder_query_support_size": 3.875,
          "decoder_belief_target_mass": 0.14285704493522644,
          "decoder_belief_target_nll": 13.206811904907227,
          "decoder_belief_top1_on_support": 0.203125,
          "mean_decoder_belief_support_size": 4.21875,
          "mean_slot_purity": 0.9999986290931702,
          "mean_slot_entropy": 2.3037249775370583e-05,
          "mean_weight_entropy": 1.9459102153778076,
          "mean_route_entropy": -1.0,
          "mean_route_accuracy": -1.0,
          "mean_target_support_size": 4.21875
        },
        {
          "model": "cyclic_transition_ladder",
          "variant": "smoke_cyclic_mixer",
          "modulus": 7,
          "observe_mod": 3,
          "observe_prob": 0.4,
          "op_family": "full",
          "query_family": "all",
          "slot_capacity": 7,
          "init_mode": "oracle",
          "transition_mode": "cyclic_mixer",
          "supervision": "full_belief",
          "slot_dim": 64,
          "hidden_dim": 128,
          "fourier_order": 8,
          "train_steps": 500,
          "length": 2,
          "query_type": "B",
          "k": 1,
          "n": 128,
          "decoder_query_target_mass": 0.7355202436447144,
          "decoder_query_target_nll": 3.3396947383880615,
          "decoder_query_top1_on_support": 0.765625,
          "mean_decoder_query_support_size": 3.875,
          "decoder_belief_target_mass": 0.3368918299674988,
          "decoder_belief_target_nll": 8.7957124710083,
          "decoder_belief_top1_on_support": 0.4296875,
          "mean_decoder_belief_support_size": 4.21875,
          "mean_slot_purity": 0.8484996557235718,
          "mean_slot_entropy": 0.22595296800136566,
          "mean_weight_entropy": 1.4873778820037842,
          "mean_route_entropy": 0.037653930485248566,
          "mean_route_accuracy": 0.8984375,
          "mean_target_support_size": 4.21875
        },
        {
          "model": "cyclic_transition_ladder",
          "variant": "smoke_cyclic_mixer",
          "modulus": 7,
          "observe_mod": 3,
          "observe_prob": 0.4,
          "op_family": "full",
          "query_family": "all",
          "slot_capacity": 7,
          "init_mode": "oracle",
          "transition_mode": "cyclic_mixer",
          "supervision": "full_belief",
          "slot_dim": 64,
          "hidden_dim": 128,
          "fourier_order": 8,
          "train_steps": 500,
          "length": 2,
          "query_type": "B",
          "k": 2,
          "n": 128,
          "decoder_query_target_mass": 0.9999688863754272,
          "decoder_query_target_nll": 1.1436047554016113,
          "decoder_query_top1_on_support": 1.0,
          "mean_decoder_query_support_size": 3.875,
          "decoder_belief_target_mass": 0.999872624874115,
          "decoder_belief_target_nll": 1.2759898900985718,
          "decoder_belief_top1_on_support": 1.0,
          "mean_decoder_belief_support_size": 4.21875,
          "mean_slot_purity": 0.8094375133514404,
          "mean_slot_entropy": 0.31659650802612305,
          "mean_weight_entropy": 1.2759504318237305,
          "mean_route_entropy": 0.03533649817109108,
          "mean_route_accuracy": 0.8880208730697632,
          "mean_target_support_size": 4.21875
        },
        {
          "model": "cyclic_transition_ladder",
          "variant": "smoke_cyclic_mixer",
          "modulus": 7,
          "observe_mod": 3,
          "observe_prob": 0.4,
          "op_family": "full",
          "query_family": "all",
          "slot_capacity": 7,
          "init_mode": "oracle",
          "transition_mode": "cyclic_mixer",
          "supervision": "full_belief",
          "slot_dim": 64,
          "hidden_dim": 128,
          "fourier_order": 8,
          "train_steps": 500,
          "length": 2,
          "query_type": "B",
          "k": 3,
          "n": 128,
          "decoder_query_target_mass": 0.9999688863754272,
          "decoder_query_target_nll": 1.1436047554016113,
          "decoder_query_top1_on_support": 1.0,
          "mean_decoder_query_support_size": 3.875,
          "decoder_belief_target_mass": 0.999872624874115,
          "decoder_belief_target_nll": 1.2759898900985718,
          "decoder_belief_top1_on_support": 1.0,
          "mean_decoder_belief_support_size": 4.21875,
          "mean_slot_purity": 0.8094375133514404,
          "mean_slot_entropy": 0.31659650802612305,
          "mean_weight_entropy": 1.2759504318237305,
          "mean_route_entropy": -1.0,
          "mean_route_accuracy": -1.0,
          "mean_target_support_size": 4.21875
        },
        {
          "model": "cyclic_transition_ladder",
          "variant": "smoke_cyclic_mixer",
          "modulus": 7,
          "observe_mod": 3,
          "observe_prob": 0.4,
          "op_family": "full",
          "query_family": "all",
          "slot_capacity": 7,
          "init_mode": "oracle",
          "transition_mode": "cyclic_mixer",
          "supervision": "full_belief",
          "slot_dim": 64,
          "hidden_dim": 128,
          "fourier_order": 8,
          "train_steps": 500,
          "length": 2,
          "query_type": "A_PLUS_B",
          "k": 0,
          "n": 128,
          "decoder_query_target_mass": 0.5758928060531616,
          "decoder_query_target_nll": 1.9459102153778076,
          "decoder_query_top1_on_support": 0.609375,
          "mean_decoder_query_support_size": 4.03125,
          "decoder_belief_target_mass": 0.12723207473754883,
          "decoder_belief_target_nll": 13.166950225830078,
          "decoder_belief_top1_on_support": 0.1875,
          "mean_decoder_belief_support_size": 4.078125,
          "mean_slot_purity": 0.9999986290931702,
          "mean_slot_entropy": 2.3037249775370583e-05,
          "mean_weight_entropy": 1.9459102153778076,
          "mean_route_entropy": -1.0,
          "mean_route_accuracy": -1.0,
          "mean_target_support_size": 4.078125
        },
        {
          "model": "cyclic_transition_ladder",
          "variant": "smoke_cyclic_mixer",
          "modulus": 7,
          "observe_mod": 3,
          "observe_prob": 0.4,
          "op_family": "full",
          "query_family": "all",
          "slot_capacity": 7,
          "init_mode": "oracle",
          "transition_mode": "cyclic_mixer",
          "supervision": "full_belief",
          "slot_dim": 64,
          "hidden_dim": 128,
          "fourier_order": 8,
          "train_steps": 500,
          "length": 2,
          "query_type": "A_PLUS_B",
          "k": 1,
          "n": 128,
          "decoder_query_target_mass": 0.6579360961914062,
          "decoder_query_target_nll": 3.758129358291626,
          "decoder_query_top1_on_support": 0.6640625,
          "mean_decoder_query_support_size": 4.03125,
          "decoder_belief_target_mass": 0.34212398529052734,
          "decoder_belief_target_nll": 8.370363235473633,
          "decoder_belief_top1_on_support": 0.3671875,
          "mean_decoder_belief_support_size": 4.078125,
          "mean_slot_purity": 0.8494381904602051,
          "mean_slot_entropy": 0.22063830494880676,
          "mean_weight_entropy": 1.4715378284454346,
          "mean_route_entropy": 0.03130786120891571,
          "mean_route_accuracy": 0.8880208730697632,
          "mean_target_support_size": 4.078125
        },
        {
          "model": "cyclic_transition_ladder",
          "variant": "smoke_cyclic_mixer",
          "modulus": 7,
          "observe_mod": 3,
          "observe_prob": 0.4,
          "op_family": "full",
          "query_family": "all",
          "slot_capacity": 7,
          "init_mode": "oracle",
          "transition_mode": "cyclic_mixer",
          "supervision": "full_belief",
          "slot_dim": 64,
          "hidden_dim": 128,
          "fourier_order": 8,
          "train_steps": 500,
          "length": 2,
          "query_type": "A_PLUS_B",
          "k": 2,
          "n": 128,
          "decoder_query_target_mass": 0.9999475479125977,
          "decoder_query_target_nll": 1.203003168106079,
          "decoder_query_top1_on_support": 1.0,
          "mean_decoder_query_support_size": 4.03125,
          "decoder_belief_target_mass": 0.9998723268508911,
          "decoder_belief_target_nll": 1.2182809114456177,
          "decoder_belief_top1_on_support": 1.0,
          "mean_decoder_belief_support_size": 4.078125,
          "mean_slot_purity": 0.7894984483718872,
          "mean_slot_entropy": 0.35043859481811523,
          "mean_weight_entropy": 1.2182152271270752,
          "mean_route_entropy": 0.038677461445331573,
          "mean_route_accuracy": 0.8802083730697632,
          "mean_target_support_size": 4.078125
        },
        {
          "model": "cyclic_transition_ladder",
          "variant": "smoke_cyclic_mixer",
          "modulus": 7,
          "observe_mod": 3,
          "observe_prob": 0.4,
          "op_family": "full",
          "query_family": "all",
          "slot_capacity": 7,
          "init_mode": "oracle",
          "transition_mode": "cyclic_mixer",
          "supervision": "full_belief",
          "slot_dim": 64,
          "hidden_dim": 128,
          "fourier_order": 8,
          "train_steps": 500,
          "length": 2,
          "query_type": "A_PLUS_B",
          "k": 3,
          "n": 128,
          "decoder_query_target_mass": 0.9999475479125977,
          "decoder_query_target_nll": 1.203003168106079,
          "decoder_query_top1_on_support": 1.0,
          "mean_decoder_query_support_size": 4.03125,
          "decoder_belief_target_mass": 0.9998723268508911,
          "decoder_belief_target_nll": 1.2182809114456177,
          "decoder_belief_top1_on_support": 1.0,
          "mean_decoder_belief_support_size": 4.078125,
          "mean_slot_purity": 0.7894984483718872,
          "mean_slot_entropy": 0.35043859481811523,
          "mean_weight_entropy": 1.2182152271270752,
          "mean_route_entropy": -1.0,
          "mean_route_accuracy": -1.0,
          "mean_target_support_size": 4.078125
        },
        {
          "model": "cyclic_transition_ladder",
          "variant": "smoke_cyclic_mixer",
          "modulus": 7,
          "observe_mod": 3,
          "observe_prob": 0.4,
          "op_family": "full",
          "query_family": "all",
          "slot_capacity": 7,
          "init_mode": "oracle",
          "transition_mode": "cyclic_mixer",
          "supervision": "full_belief",
          "slot_dim": 64,
          "hidden_dim": 128,
          "fourier_order": 8,
          "train_steps": 500,
          "length": 2,
          "query_type": "A_MINUS_B",
          "k": 0,
          "n": 128,
          "decoder_query_target_mass": 0.5312498807907104,
          "decoder_query_target_nll": 11.628198623657227,
          "decoder_query_top1_on_support": 0.53125,
          "mean_decoder_query_support_size": 3.046875,
          "decoder_belief_target_mass": 0.11607136577367783,
          "decoder_belief_target_nll": 13.574108123779297,
          "decoder_belief_top1_on_support": 0.171875,
          "mean_decoder_belief_support_size": 4.2265625,
          "mean_slot_purity": 0.9999986290931702,
          "mean_slot_entropy": 2.3037249775370583e-05,
          "mean_weight_entropy": 1.9459102153778076,
          "mean_route_entropy": -1.0,
          "mean_route_accuracy": -1.0,
          "mean_target_support_size": 4.2265625
        },
        {
          "model": "cyclic_transition_ladder",
          "variant": "smoke_cyclic_mixer",
          "modulus": 7,
          "observe_mod": 3,
          "observe_prob": 0.4,
          "op_family": "full",
          "query_family": "all",
          "slot_capacity": 7,
          "init_mode": "oracle",
          "transition_mode": "cyclic_mixer",
          "supervision": "full_belief",
          "slot_dim": 64,
          "hidden_dim": 128,
          "fourier_order": 8,
          "train_steps": 500,
          "length": 2,
          "query_type": "A_MINUS_B",
          "k": 1,
          "n": 128,
          "decoder_query_target_mass": 0.6818940043449402,
          "decoder_query_target_nll": 4.879995822906494,
          "decoder_query_top1_on_support": 0.6796875,
          "mean_decoder_query_support_size": 3.046875,
          "decoder_belief_target_mass": 0.3561883270740509,
          "decoder_belief_target_nll": 7.931352615356445,
          "decoder_belief_top1_on_support": 0.359375,
          "mean_decoder_belief_support_size": 4.2265625,
          "mean_slot_purity": 0.8777118921279907,
          "mean_slot_entropy": 0.179824560880661,
          "mean_weight_entropy": 1.5630743503570557,
          "mean_route_entropy": 0.04031972587108612,
          "mean_route_accuracy": 0.890625,
          "mean_target_support_size": 4.2265625
        },
        {
          "model": "cyclic_transition_ladder",
          "variant": "smoke_cyclic_mixer",
          "modulus": 7,
          "observe_mod": 3,
          "observe_prob": 0.4,
          "op_family": "full",
          "query_family": "all",
          "slot_capacity": 7,
          "init_mode": "oracle",
          "transition_mode": "cyclic_mixer",
          "supervision": "full_belief",
          "slot_dim": 64,
          "hidden_dim": 128,
          "fourier_order": 8,
          "train_steps": 500,
          "length": 2,
          "query_type": "A_MINUS_B",
          "k": 2,
          "n": 128,
          "decoder_query_target_mass": 0.9999211430549622,
          "decoder_query_target_nll": 0.7439091801643372,
          "decoder_query_top1_on_support": 1.0,
          "mean_decoder_query_support_size": 3.046875,
          "decoder_belief_target_mass": 0.9998705387115479,
          "decoder_belief_target_nll": 1.2708604335784912,
          "decoder_belief_top1_on_support": 1.0,
          "mean_decoder_belief_support_size": 4.2265625,
          "mean_slot_purity": 0.8215991258621216,
          "mean_slot_entropy": 0.30015742778778076,
          "mean_weight_entropy": 1.2708497047424316,
          "mean_route_entropy": 0.03675951808691025,
          "mean_route_accuracy": 0.9010416865348816,
          "mean_target_support_size": 4.2265625
        },
        {
          "model": "cyclic_transition_ladder",
          "variant": "smoke_cyclic_mixer",
          "modulus": 7,
          "observe_mod": 3,
          "observe_prob": 0.4,
          "op_family": "full",
          "query_family": "all",
          "slot_capacity": 7,
          "init_mode": "oracle",
          "transition_mode": "cyclic_mixer",
          "supervision": "full_belief",
          "slot_dim": 64,
          "hidden_dim": 128,
          "fourier_order": 8,
          "train_steps": 500,
          "length": 2,
          "query_type": "A_MINUS_B",
          "k": 3,
          "n": 128,
          "decoder_query_target_mass": 0.9999211430549622,
          "decoder_query_target_nll": 0.7439091801643372,
          "decoder_query_top1_on_support": 1.0,
          "mean_decoder_query_support_size": 3.046875,
          "decoder_belief_target_mass": 0.9998705387115479,
          "decoder_belief_target_nll": 1.2708604335784912,
          "decoder_belief_top1_on_support": 1.0,
          "mean_decoder_belief_support_size": 4.2265625,
          "mean_slot_purity": 0.8215991258621216,
          "mean_slot_entropy": 0.30015742778778076,
          "mean_weight_entropy": 1.2708497047424316,
          "mean_route_entropy": -1.0,
          "mean_route_accuracy": -1.0,
          "mean_target_support_size": 4.2265625
        },
        {
          "model": "cyclic_transition_ladder",
          "variant": "smoke_cyclic_mixer",
          "modulus": 7,
          "observe_mod": 3,
          "observe_prob": 0.4,
          "op_family": "full",
          "query_family": "all",
          "slot_capacity": 7,
          "init_mode": "oracle",
          "transition_mode": "cyclic_mixer",
          "supervision": "full_belief",
          "slot_dim": 64,
          "hidden_dim": 128,
          "fourier_order": 8,
          "train_steps": 500,
          "length": 3,
          "query_type": "A",
          "k": 0,
          "n": 128,
          "decoder_query_target_mass": 0.4374999701976776,
          "decoder_query_target_nll": 1.9459102153778076,
          "decoder_query_top1_on_support": 0.3984375,
          "mean_decoder_query_support_size": 3.0625,
          "decoder_belief_target_mass": 0.0658482164144516,
          "decoder_belief_target_nll": 15.140100479125977,
          "decoder_belief_top1_on_support": 0.125,
          "mean_decoder_belief_support_size": 3.359375,
          "mean_slot_purity": 0.9999986290931702,
          "mean_slot_entropy": 2.3037249775370583e-05,
          "mean_weight_entropy": 1.9459102153778076,
          "mean_route_entropy": -1.0,
          "mean_route_accuracy": -1.0,
          "mean_target_support_size": 3.359375
        },
        {
          "model": "cyclic_transition_ladder",
          "variant": "smoke_cyclic_mixer",
          "modulus": 7,
          "observe_mod": 3,
          "observe_prob": 0.4,
          "op_family": "full",
          "query_family": "all",
          "slot_capacity": 7,
          "init_mode": "oracle",
          "transition_mode": "cyclic_mixer",
          "supervision": "full_belief",
          "slot_dim": 64,
          "hidden_dim": 128,
          "fourier_order": 8,
          "train_steps": 500,
          "length": 3,
          "query_type": "A",
          "k": 1,
          "n": 128,
          "decoder_query_target_mass": 0.5554220676422119,
          "decoder_query_target_nll": 3.8128395080566406,
          "decoder_query_top1_on_support": 0.546875,
          "mean_decoder_query_support_size": 3.0625,
          "decoder_belief_target_mass": 0.19850113987922668,
          "decoder_belief_target_nll": 10.953351974487305,
          "decoder_belief_top1_on_support": 0.1953125,
          "mean_decoder_belief_support_size": 3.359375,
          "mean_slot_purity": 0.8566860556602478,
          "mean_slot_entropy": 0.21535852551460266,
          "mean_weight_entropy": 1.5199061632156372,
          "mean_route_entropy": 0.04087726026773453,
          "mean_route_accuracy": 0.9010416865348816,
          "mean_target_support_size": 3.359375
        },
        {
          "model": "cyclic_transition_ladder",
          "variant": "smoke_cyclic_mixer",
          "modulus": 7,
          "observe_mod": 3,
          "observe_prob": 0.4,
          "op_family": "full",
          "query_family": "all",
          "slot_capacity": 7,
          "init_mode": "oracle",
          "transition_mode": "cyclic_mixer",
          "supervision": "full_belief",
          "slot_dim": 64,
          "hidden_dim": 128,
          "fourier_order": 8,
          "train_steps": 500,
          "length": 3,
          "query_type": "A",
          "k": 2,
          "n": 128,
          "decoder_query_target_mass": 0.7310171127319336,
          "decoder_query_target_nll": 3.2301342487335205,
          "decoder_query_top1_on_support": 0.6796875,
          "mean_decoder_query_support_size": 3.0625,
          "decoder_belief_target_mass": 0.3677597641944885,
          "decoder_belief_target_nll": 7.983285903930664,
          "decoder_belief_top1_on_support": 0.3515625,
          "mean_decoder_belief_support_size": 3.359375,
          "mean_slot_purity": 0.7968101501464844,
          "mean_slot_entropy": 0.34729844331741333,
          "mean_weight_entropy": 1.1966265439987183,
          "mean_route_entropy": 0.03646457567811012,
          "mean_route_accuracy": 0.8984375,
          "mean_target_support_size": 3.359375
        },
        {
          "model": "cyclic_transition_ladder",
          "variant": "smoke_cyclic_mixer",
          "modulus": 7,
          "observe_mod": 3,
          "observe_prob": 0.4,
          "op_family": "full",
          "query_family": "all",
          "slot_capacity": 7,
          "init_mode": "oracle",
          "transition_mode": "cyclic_mixer",
          "supervision": "full_belief",
          "slot_dim": 64,
          "hidden_dim": 128,
          "fourier_order": 8,
          "train_steps": 500,
          "length": 3,
          "query_type": "A",
          "k": 3,
          "n": 128,
          "decoder_query_target_mass": 0.9999369978904724,
          "decoder_query_target_nll": 0.9006429314613342,
          "decoder_query_top1_on_support": 1.0,
          "mean_decoder_query_support_size": 3.0625,
          "decoder_belief_target_mass": 0.9997962117195129,
          "decoder_belief_target_nll": 1.0178704261779785,
          "decoder_belief_top1_on_support": 1.0,
          "mean_decoder_belief_support_size": 3.359375,
          "mean_slot_purity": 0.7719928026199341,
          "mean_slot_entropy": 0.42912161350250244,
          "mean_weight_entropy": 1.0178945064544678,
          "mean_route_entropy": 0.03442591428756714,
          "mean_route_accuracy": 0.8776041865348816,
          "mean_target_support_size": 3.359375
        },
        {
          "model": "cyclic_transition_ladder",
          "variant": "smoke_cyclic_mixer",
          "modulus": 7,
          "observe_mod": 3,
          "observe_prob": 0.4,
          "op_family": "full",
          "query_family": "all",
          "slot_capacity": 7,
          "init_mode": "oracle",
          "transition_mode": "cyclic_mixer",
          "supervision": "full_belief",
          "slot_dim": 64,
          "hidden_dim": 128,
          "fourier_order": 8,
          "train_steps": 500,
          "length": 3,
          "query_type": "B",
          "k": 0,
          "n": 128,
          "decoder_query_target_mass": 0.3660714030265808,
          "decoder_query_target_nll": 1.9459102153778076,
          "decoder_query_top1_on_support": 0.3671875,
          "mean_decoder_query_support_size": 2.5625,
          "decoder_belief_target_mass": 0.07477675378322601,
          "decoder_belief_target_nll": 14.422590255737305,
          "decoder_belief_top1_on_support": 0.1015625,
          "mean_decoder_belief_support_size": 3.03125,
          "mean_slot_purity": 0.9999986290931702,
          "mean_slot_entropy": 2.3037249775370583e-05,
          "mean_weight_entropy": 1.9459102153778076,
          "mean_route_entropy": -1.0,
          "mean_route_accuracy": -1.0,
          "mean_target_support_size": 3.03125
        },
        {
          "model": "cyclic_transition_ladder",
          "variant": "smoke_cyclic_mixer",
          "modulus": 7,
          "observe_mod": 3,
          "observe_prob": 0.4,
          "op_family": "full",
          "query_family": "all",
          "slot_capacity": 7,
          "init_mode": "oracle",
          "transition_mode": "cyclic_mixer",
          "supervision": "full_belief",
          "slot_dim": 64,
          "hidden_dim": 128,
          "fourier_order": 8,
          "train_steps": 500,
          "length": 3,
          "query_type": "B",
          "k": 1,
          "n": 128,
          "decoder_query_target_mass": 0.4706054925918579,
          "decoder_query_target_nll": 4.934222221374512,
          "decoder_query_top1_on_support": 0.46875,
          "mean_decoder_query_support_size": 2.5625,
          "decoder_belief_target_mass": 0.2103562355041504,
          "decoder_belief_target_nll": 11.005644798278809,
          "decoder_belief_top1_on_support": 0.1875,
          "mean_decoder_belief_support_size": 3.03125,
          "mean_slot_purity": 0.8438546657562256,
          "mean_slot_entropy": 0.22841179370880127,
          "mean_weight_entropy": 1.45196533203125,
          "mean_route_entropy": 0.04012569785118103,
          "mean_route_accuracy": 0.9010416865348816,
          "mean_target_support_size": 3.03125
        },
        {
          "model": "cyclic_transition_ladder",
          "variant": "smoke_cyclic_mixer",
          "modulus": 7,
          "observe_mod": 3,
          "observe_prob": 0.4,
          "op_family": "full",
          "query_family": "all",
          "slot_capacity": 7,
          "init_mode": "oracle",
          "transition_mode": "cyclic_mixer",
          "supervision": "full_belief",
          "slot_dim": 64,
          "hidden_dim": 128,
          "fourier_order": 8,
          "train_steps": 500,
          "length": 3,
          "query_type": "B",
          "k": 2,
          "n": 128,
          "decoder_query_target_mass": 0.692685067653656,
          "decoder_query_target_nll": 3.264869213104248,
          "decoder_query_top1_on_support": 0.6875,
          "mean_decoder_query_support_size": 2.5625,
          "decoder_belief_target_mass": 0.3930370807647705,
          "decoder_belief_target_nll": 7.314023971557617,
          "decoder_belief_top1_on_support": 0.3984375,
          "mean_decoder_belief_support_size": 3.03125,
          "mean_slot_purity": 0.7858361005783081,
          "mean_slot_entropy": 0.36682918667793274,
          "mean_weight_entropy": 1.145761251449585,
          "mean_route_entropy": 0.038544610142707825,
          "mean_route_accuracy": 0.9114583730697632,
          "mean_target_support_size": 3.03125
        },
        {
          "model": "cyclic_transition_ladder",
          "variant": "smoke_cyclic_mixer",
          "modulus": 7,
          "observe_mod": 3,
          "observe_prob": 0.4,
          "op_family": "full",
          "query_family": "all",
          "slot_capacity": 7,
          "init_mode": "oracle",
          "transition_mode": "cyclic_mixer",
          "supervision": "full_belief",
          "slot_dim": 64,
          "hidden_dim": 128,
          "fourier_order": 8,
          "train_steps": 500,
          "length": 3,
          "query_type": "B",
          "k": 3,
          "n": 128,
          "decoder_query_target_mass": 0.9999251365661621,
          "decoder_query_target_nll": 0.7442592978477478,
          "decoder_query_top1_on_support": 1.0,
          "mean_decoder_query_support_size": 2.5625,
          "decoder_belief_target_mass": 0.9998071789741516,
          "decoder_belief_target_nll": 0.9370309114456177,
          "decoder_belief_top1_on_support": 1.0,
          "mean_decoder_belief_support_size": 3.03125,
          "mean_slot_purity": 0.7496495246887207,
          "mean_slot_entropy": 0.48816680908203125,
          "mean_weight_entropy": 0.9371389746665955,
          "mean_route_entropy": 0.037614405155181885,
          "mean_route_accuracy": 0.8984375,
          "mean_target_support_size": 3.03125
        },
        {
          "model": "cyclic_transition_ladder",
          "variant": "smoke_cyclic_mixer",
          "modulus": 7,
          "observe_mod": 3,
          "observe_prob": 0.4,
          "op_family": "full",
          "query_family": "all",
          "slot_capacity": 7,
          "init_mode": "oracle",
          "transition_mode": "cyclic_mixer",
          "supervision": "full_belief",
          "slot_dim": 64,
          "hidden_dim": 128,
          "fourier_order": 8,
          "train_steps": 500,
          "length": 3,
          "query_type": "A_PLUS_B",
          "k": 0,
          "n": 128,
          "decoder_query_target_mass": 0.4084821343421936,
          "decoder_query_target_nll": 1.9459102153778076,
          "decoder_query_top1_on_support": 0.453125,
          "mean_decoder_query_support_size": 2.859375,
          "decoder_belief_target_mass": 0.07477675378322601,
          "decoder_belief_target_nll": 14.519397735595703,
          "decoder_belief_top1_on_support": 0.1171875,
          "mean_decoder_belief_support_size": 2.984375,
          "mean_slot_purity": 0.9999986290931702,
          "mean_slot_entropy": 2.3037249775370583e-05,
          "mean_weight_entropy": 1.9459102153778076,
          "mean_route_entropy": -1.0,
          "mean_route_accuracy": -1.0,
          "mean_target_support_size": 2.984375
        },
        {
          "model": "cyclic_transition_ladder",
          "variant": "smoke_cyclic_mixer",
          "modulus": 7,
          "observe_mod": 3,
          "observe_prob": 0.4,
          "op_family": "full",
          "query_family": "all",
          "slot_capacity": 7,
          "init_mode": "oracle",
          "transition_mode": "cyclic_mixer",
          "supervision": "full_belief",
          "slot_dim": 64,
          "hidden_dim": 128,
          "fourier_order": 8,
          "train_steps": 500,
          "length": 3,
          "query_type": "A_PLUS_B",
          "k": 1,
          "n": 128,
          "decoder_query_target_mass": 0.4540975093841553,
          "decoder_query_target_nll": 4.519122123718262,
          "decoder_query_top1_on_support": 0.4609375,
          "mean_decoder_query_support_size": 2.859375,
          "decoder_belief_target_mass": 0.1832653433084488,
          "decoder_belief_target_nll": 11.124882698059082,
          "decoder_belief_top1_on_support": 0.1953125,
          "mean_decoder_belief_support_size": 2.984375,
          "mean_slot_purity": 0.837348222732544,
          "mean_slot_entropy": 0.24129199981689453,
          "mean_weight_entropy": 1.448229432106018,
          "mean_route_entropy": 0.039275504648685455,
          "mean_route_accuracy": 0.9036458730697632,
          "mean_target_support_size": 2.984375
        },
        {
          "model": "cyclic_transition_ladder",
          "variant": "smoke_cyclic_mixer",
          "modulus": 7,
          "observe_mod": 3,
          "observe_prob": 0.4,
          "op_family": "full",
          "query_family": "all",
          "slot_capacity": 7,
          "init_mode": "oracle",
          "transition_mode": "cyclic_mixer",
          "supervision": "full_belief",
          "slot_dim": 64,
          "hidden_dim": 128,
          "fourier_order": 8,
          "train_steps": 500,
          "length": 3,
          "query_type": "A_PLUS_B",
          "k": 2,
          "n": 128,
          "decoder_query_target_mass": 0.542128324508667,
          "decoder_query_target_nll": 4.712558746337891,
          "decoder_query_top1_on_support": 0.6171875,
          "mean_decoder_query_support_size": 2.859375,
          "decoder_belief_target_mass": 0.36822453141212463,
          "decoder_belief_target_nll": 7.850386619567871,
          "decoder_belief_top1_on_support": 0.453125,
          "mean_decoder_belief_support_size": 2.984375,
          "mean_slot_purity": 0.763562023639679,
          "mean_slot_entropy": 0.4212270677089691,
          "mean_weight_entropy": 1.0989538431167603,
          "mean_route_entropy": 0.03556730970740318,
          "mean_route_accuracy": 0.9036458730697632,
          "mean_target_support_size": 2.984375
        },
        {
          "model": "cyclic_transition_ladder",
          "variant": "smoke_cyclic_mixer",
          "modulus": 7,
          "observe_mod": 3,
          "observe_prob": 0.4,
          "op_family": "full",
          "query_family": "all",
          "slot_capacity": 7,
          "init_mode": "oracle",
          "transition_mode": "cyclic_mixer",
          "supervision": "full_belief",
          "slot_dim": 64,
          "hidden_dim": 128,
          "fourier_order": 8,
          "train_steps": 500,
          "length": 3,
          "query_type": "A_PLUS_B",
          "k": 3,
          "n": 128,
          "decoder_query_target_mass": 0.9998795986175537,
          "decoder_query_target_nll": 0.8651937246322632,
          "decoder_query_top1_on_support": 1.0,
          "mean_decoder_query_support_size": 2.859375,
          "decoder_belief_target_mass": 0.9998166561126709,
          "decoder_belief_target_nll": 0.9128274917602539,
          "decoder_belief_top1_on_support": 1.0,
          "mean_decoder_belief_support_size": 2.984375,
          "mean_slot_purity": 0.7282586097717285,
          "mean_slot_entropy": 0.5304306745529175,
          "mean_weight_entropy": 0.9128948450088501,
          "mean_route_entropy": 0.0373956635594368,
          "mean_route_accuracy": 0.90625,
          "mean_target_support_size": 2.984375
        },
        {
          "model": "cyclic_transition_ladder",
          "variant": "smoke_cyclic_mixer",
          "modulus": 7,
          "observe_mod": 3,
          "observe_prob": 0.4,
          "op_family": "full",
          "query_family": "all",
          "slot_capacity": 7,
          "init_mode": "oracle",
          "transition_mode": "cyclic_mixer",
          "supervision": "full_belief",
          "slot_dim": 64,
          "hidden_dim": 128,
          "fourier_order": 8,
          "train_steps": 500,
          "length": 3,
          "query_type": "A_MINUS_B",
          "k": 0,
          "n": 128,
          "decoder_query_target_mass": 0.5156247615814209,
          "decoder_query_target_nll": 11.574100494384766,
          "decoder_query_top1_on_support": 0.515625,
          "mean_decoder_query_support_size": 2.7109375,
          "decoder_belief_target_mass": 0.08928567171096802,
          "decoder_belief_target_nll": 13.520009994506836,
          "decoder_belief_top1_on_support": 0.1328125,
          "mean_decoder_belief_support_size": 3.1796875,
          "mean_slot_purity": 0.9999986290931702,
          "mean_slot_entropy": 2.3037249775370583e-05,
          "mean_weight_entropy": 1.9459102153778076,
          "mean_route_entropy": -1.0,
          "mean_route_accuracy": -1.0,
          "mean_target_support_size": 3.1796875
        },
        {
          "model": "cyclic_transition_ladder",
          "variant": "smoke_cyclic_mixer",
          "modulus": 7,
          "observe_mod": 3,
          "observe_prob": 0.4,
          "op_family": "full",
          "query_family": "all",
          "slot_capacity": 7,
          "init_mode": "oracle",
          "transition_mode": "cyclic_mixer",
          "supervision": "full_belief",
          "slot_dim": 64,
          "hidden_dim": 128,
          "fourier_order": 8,
          "train_steps": 500,
          "length": 3,
          "query_type": "A_MINUS_B",
          "k": 1,
          "n": 128,
          "decoder_query_target_mass": 0.5100405216217041,
          "decoder_query_target_nll": 7.032371520996094,
          "decoder_query_top1_on_support": 0.5,
          "mean_decoder_query_support_size": 2.7109375,
          "decoder_belief_target_mass": 0.2449864000082016,
          "decoder_belief_target_nll": 10.888083457946777,
          "decoder_belief_top1_on_support": 0.2890625,
          "mean_decoder_belief_support_size": 3.1796875,
          "mean_slot_purity": 0.8323264122009277,
          "mean_slot_entropy": 0.2459137737751007,
          "mean_weight_entropy": 1.419152021408081,
          "mean_route_entropy": 0.03898832947015762,
          "mean_route_accuracy": 0.9114583730697632,
          "mean_target_support_size": 3.1796875
        },
        {
          "model": "cyclic_transition_ladder",
          "variant": "smoke_cyclic_mixer",
          "modulus": 7,
          "observe_mod": 3,
          "observe_prob": 0.4,
          "op_family": "full",
          "query_family": "all",
          "slot_capacity": 7,
          "init_mode": "oracle",
          "transition_mode": "cyclic_mixer",
          "supervision": "full_belief",
          "slot_dim": 64,
          "hidden_dim": 128,
          "fourier_order": 8,
          "train_steps": 500,
          "length": 3,
          "query_type": "A_MINUS_B",
          "k": 2,
          "n": 128,
          "decoder_query_target_mass": 0.6050550937652588,
          "decoder_query_target_nll": 5.098701477050781,
          "decoder_query_top1_on_support": 0.6015625,
          "mean_decoder_query_support_size": 2.7109375,
          "decoder_belief_target_mass": 0.4020967483520508,
          "decoder_belief_target_nll": 7.8017048835754395,
          "decoder_belief_top1_on_support": 0.421875,
          "mean_decoder_belief_support_size": 3.1796875,
          "mean_slot_purity": 0.7722219228744507,
          "mean_slot_entropy": 0.3897216320037842,
          "mean_weight_entropy": 1.120895266532898,
          "mean_route_entropy": 0.03316420316696167,
          "mean_route_accuracy": 0.9010416865348816,
          "mean_target_support_size": 3.1796875
        },
        {
          "model": "cyclic_transition_ladder",
          "variant": "smoke_cyclic_mixer",
          "modulus": 7,
          "observe_mod": 3,
          "observe_prob": 0.4,
          "op_family": "full",
          "query_family": "all",
          "slot_capacity": 7,
          "init_mode": "oracle",
          "transition_mode": "cyclic_mixer",
          "supervision": "full_belief",
          "slot_dim": 64,
          "hidden_dim": 128,
          "fourier_order": 8,
          "train_steps": 500,
          "length": 3,
          "query_type": "A_MINUS_B",
          "k": 3,
          "n": 128,
          "decoder_query_target_mass": 0.9998821020126343,
          "decoder_query_target_nll": 0.705075740814209,
          "decoder_query_top1_on_support": 1.0,
          "mean_decoder_query_support_size": 2.7109375,
          "decoder_belief_target_mass": 0.9998207092285156,
          "decoder_belief_target_nll": 0.9646987915039062,
          "decoder_belief_top1_on_support": 1.0,
          "mean_decoder_belief_support_size": 3.1796875,
          "mean_slot_purity": 0.7427268028259277,
          "mean_slot_entropy": 0.48876333236694336,
          "mean_weight_entropy": 0.9647326469421387,
          "mean_route_entropy": 0.03563893958926201,
          "mean_route_accuracy": 0.9010416865348816,
          "mean_target_support_size": 3.1796875
        }
      ]
    }
  ],
  "checkpoints": [
    "large_artifacts/cyclic_transition_ladder/checkpoints/smoke_cyclic_mixer/checkpoint_final.pt"
  ],
  "num_parameters": 36047
}