{
  "metadata": {
    "python": "3.12.3",
    "platform": "Linux-6.8.0-62-generic-x86_64-with-glibc2.39",
    "torch": "2.8.0+cu128",
    "cuda_available": true,
    "gpu_name": "NVIDIA RTX 6000 Ada Generation",
    "args": {
      "run_name": "pilot_iterative_k1_sparse_s96",
      "compiler_checkpoint": "large_artifacts/qwen_iterative_repair_policy/checkpoints/fixed_compiler_step00800",
      "train_examples": 96,
      "val_examples": 48,
      "eval_examples": 64,
      "eval_pairs": 64,
      "train_length": 24,
      "eval_length": 24,
      "train_template_modes": "standard,paraphrase",
      "repair_topk": 3,
      "repair_max_edits": 2,
      "repair_max_pair_arg_slots": 24,
      "qwen_batch_size": 8,
      "max_length": 384,
      "editor_d_model": 128,
      "editor_layers": 2,
      "editor_heads": 4,
      "editor_ff_mult": 4,
      "editor_dropout": 0.05,
      "editor_epochs": 6,
      "editor_lr": 0.001,
      "editor_weight_decay": 0.0001,
      "editor_target_mode": "oracle_or_gold",
      "changed_slot_weight": 4.0,
      "base_copy_logit_bias": 5.0,
      "edit_gate_loss_weight": 1.0,
      "edit_value_loss_weight": 1.0,
      "edit_gate_pos_weight": 8.0,
      "unchanged_value_loss_weight": 0.05,
      "edit_threshold_grid": "0.2,0.3,0.4,0.5,0.6,0.7,0.8",
      "train_select_k_values": "1,2,3",
      "eval_k_values": "0,1,2,3",
      "max_edits_per_iteration": 1,
      "augment_cases_per_group": 2,
      "augment_top_pool": 16,
      "augment_repairable_only": false,
      "include_target_noedit_case": true,
      "stats_candidate_traces": 6,
      "init_loss_weight": 1.0,
      "op_loss_weight": 1.0,
      "arg_loss_weight": 1.0,
      "grad_clip": 1.0,
      "seed": 202,
      "eval_seed": 202001
    },
    "compiler_args": {
      "model_id": "Qwen/Qwen3-4B",
      "modulus": 97,
      "max_steps": 24,
      "head_width": 768,
      "rank_temperature": 1.0,
      "arg_reader_mode": "after_op",
      "arg_window": 8,
      "arg_distance_temperature": 2.0
    },
    "feature_names": [
      "bias",
      "candidate_prior",
      "prior_delta",
      "candidate_prior_per_slot",
      "prior_delta_per_edit",
      "base_prior",
      "edit_count",
      "changed",
      "init_edit",
      "op_edit_count",
      "arg_edit_count",
      "edit_fraction",
      "first_edit_pos_norm",
      "last_edit_pos_norm",
      "mean_edit_pos_norm",
      "edit_span_norm",
      "mean_new_rank",
      "max_new_rank",
      "mean_changed_margin",
      "min_changed_margin",
      "mean_changed_entropy",
      "mean_logp_delta",
      "min_logp_delta",
      "max_logp_delta",
      "base_init_margin",
      "base_op_margin_mean",
      "base_op_margin_min",
      "base_arg_margin_mean",
      "base_arg_margin_min",
      "base_op_entropy_mean",
      "base_arg_entropy_mean",
      "candidate_answer_norm",
      "candidate_final_soft_logp",
      "candidate_state_soft_logp_mean",
      "candidate_state_soft_logp_min",
      "base_final_soft_logp",
      "base_state_soft_logp_mean",
      "answer_equals_base",
      "state_same_fraction_as_base",
      "state_prefix_same_as_base",
      "mean_old_value_norm",
      "mean_new_value_norm",
      "mean_abs_value_delta_norm",
      "arg_new_value_mean_norm",
      "arg_abs_delta_mean_norm",
      "op_new_add_frac",
      "op_new_sub_frac",
      "op_new_mul_frac",
      "candidate_add_frac",
      "candidate_sub_frac",
      "candidate_mul_frac",
      "candidate_arg_mean_norm",
      "candidate_arg_std_norm",
      "base_add_frac",
      "base_sub_frac",
      "base_mul_frac"
    ],
    "trace_feature_names": [
      "is_global",
      "is_step",
      "active",
      "step_pos_norm",
      "candidate_prior",
      "prior_delta",
      "edit_count",
      "init_value_norm",
      "init_changed",
      "answer_norm",
      "final_soft_logp",
      "base_final_soft_logp",
      "answer_equals_base",
      "op_add",
      "op_sub",
      "op_mul",
      "arg_norm",
      "state_norm",
      "state_soft_logp",
      "base_op_add",
      "base_op_sub",
      "base_op_mul",
      "base_arg_norm",
      "base_state_norm",
      "base_state_soft_logp",
      "op_changed",
      "arg_changed",
      "state_same_as_base",
      "op_logp",
      "arg_logp",
      "op_logp_delta",
      "arg_logp_delta",
      "op_margin",
      "arg_margin",
      "op_entropy",
      "arg_entropy"
    ],
    "editor_type": "iterative_sparse_edit_policy",
    "best_editor_epoch": 3,
    "best_editor_threshold": 0.8,
    "best_editor_k": 1,
    "train_seconds": 580.8568215370178
  },
  "metrics": {
    "train_len24_k0": {
      "n": 96.0,
      "k_steps": 0.0,
      "threshold": 0.8,
      "oracle_found_fraction": 0.84375,
      "avg_candidates": 1299.0,
      "avg_positive_candidates": 1.1666666666666667,
      "base_executor_accuracy": 0.22916666666666666,
      "base_program_exact": 0.22916666666666666,
      "base_state_prefix_fraction": 0.5507812500000001,
      "base_changed_fraction": 0.0,
      "base_avg_edits": 0.0,
      "base_in_candidate_set_fraction": 1.0,
      "oracle_executor_accuracy": 0.84375,
      "oracle_program_exact": 0.84375,
      "oracle_state_prefix_fraction": 0.8832465277777777,
      "oracle_changed_fraction": 0.6145833333333334,
      "oracle_avg_edits": 0.8229166666666666,
      "oracle_in_candidate_set_fraction": 1.0,
      "iter_executor_accuracy": 0.22916666666666666,
      "iter_program_exact": 0.22916666666666666,
      "iter_state_prefix_fraction": 0.5507812500000001,
      "iter_changed_fraction": 0.0,
      "iter_avg_edits": 0.0,
      "iter_in_candidate_set_fraction": 1.0,
      "iter_steps_used_mean": 0.0,
      "iter_total_iter_edits_mean": 0.0,
      "iter_oracle_gap_recovered": 0.0
    },
    "train_len24_k1": {
      "n": 96.0,
      "k_steps": 1.0,
      "threshold": 0.8,
      "oracle_found_fraction": 0.84375,
      "avg_candidates": 1299.0,
      "avg_positive_candidates": 1.1666666666666667,
      "base_executor_accuracy": 0.22916666666666666,
      "base_program_exact": 0.22916666666666666,
      "base_state_prefix_fraction": 0.5507812500000001,
      "base_changed_fraction": 0.0,
      "base_avg_edits": 0.0,
      "base_in_candidate_set_fraction": 1.0,
      "oracle_executor_accuracy": 0.84375,
      "oracle_program_exact": 0.84375,
      "oracle_state_prefix_fraction": 0.8832465277777777,
      "oracle_changed_fraction": 0.6145833333333334,
      "oracle_avg_edits": 0.8229166666666666,
      "oracle_in_candidate_set_fraction": 1.0,
      "iter_executor_accuracy": 0.21875,
      "iter_program_exact": 0.21875,
      "iter_state_prefix_fraction": 0.5160590277777779,
      "iter_changed_fraction": 0.6041666666666666,
      "iter_avg_edits": 0.6041666666666666,
      "iter_in_candidate_set_fraction": 0.53125,
      "iter_steps_used_mean": 0.6041666666666666,
      "iter_total_iter_edits_mean": 0.6041666666666666,
      "iter_oracle_gap_recovered": -0.016949152542372864
    },
    "train_len24_k2": {
      "n": 96.0,
      "k_steps": 2.0,
      "threshold": 0.8,
      "oracle_found_fraction": 0.84375,
      "avg_candidates": 1299.0,
      "avg_positive_candidates": 1.1666666666666667,
      "base_executor_accuracy": 0.22916666666666666,
      "base_program_exact": 0.22916666666666666,
      "base_state_prefix_fraction": 0.5507812500000001,
      "base_changed_fraction": 0.0,
      "base_avg_edits": 0.0,
      "base_in_candidate_set_fraction": 1.0,
      "oracle_executor_accuracy": 0.84375,
      "oracle_program_exact": 0.84375,
      "oracle_state_prefix_fraction": 0.8832465277777777,
      "oracle_changed_fraction": 0.6145833333333334,
      "oracle_avg_edits": 0.8229166666666666,
      "oracle_in_candidate_set_fraction": 1.0,
      "iter_executor_accuracy": 0.23958333333333334,
      "iter_program_exact": 0.23958333333333334,
      "iter_state_prefix_fraction": 0.5269097222222223,
      "iter_changed_fraction": 0.6041666666666666,
      "iter_avg_edits": 0.6458333333333334,
      "iter_in_candidate_set_fraction": 0.53125,
      "iter_steps_used_mean": 1.1145833333333333,
      "iter_total_iter_edits_mean": 1.1145833333333333,
      "iter_oracle_gap_recovered": 0.01694915254237291
    },
    "train_len24_k3": {
      "n": 96.0,
      "k_steps": 3.0,
      "threshold": 0.8,
      "oracle_found_fraction": 0.84375,
      "avg_candidates": 1299.0,
      "avg_positive_candidates": 1.1666666666666667,
      "base_executor_accuracy": 0.22916666666666666,
      "base_program_exact": 0.22916666666666666,
      "base_state_prefix_fraction": 0.5507812500000001,
      "base_changed_fraction": 0.0,
      "base_avg_edits": 0.0,
      "base_in_candidate_set_fraction": 1.0,
      "oracle_executor_accuracy": 0.84375,
      "oracle_program_exact": 0.84375,
      "oracle_state_prefix_fraction": 0.8832465277777777,
      "oracle_changed_fraction": 0.6145833333333334,
      "oracle_avg_edits": 0.8229166666666666,
      "oracle_in_candidate_set_fraction": 1.0,
      "iter_executor_accuracy": 0.23958333333333334,
      "iter_program_exact": 0.23958333333333334,
      "iter_state_prefix_fraction": 0.5269097222222223,
      "iter_changed_fraction": 0.6041666666666666,
      "iter_avg_edits": 0.6458333333333334,
      "iter_in_candidate_set_fraction": 0.5104166666666666,
      "iter_steps_used_mean": 1.15625,
      "iter_total_iter_edits_mean": 1.15625,
      "iter_oracle_gap_recovered": 0.01694915254237291
    },
    "val_len24_k0": {
      "n": 48.0,
      "k_steps": 0.0,
      "threshold": 0.8,
      "oracle_found_fraction": 0.875,
      "avg_candidates": 1299.0,
      "avg_positive_candidates": 1.125,
      "base_executor_accuracy": 0.2708333333333333,
      "base_program_exact": 0.2708333333333333,
      "base_state_prefix_fraction": 0.5989583333333334,
      "base_changed_fraction": 0.0,
      "base_avg_edits": 0.0,
      "base_in_candidate_set_fraction": 1.0,
      "oracle_executor_accuracy": 0.875,
      "oracle_program_exact": 0.875,
      "oracle_state_prefix_fraction": 0.9027777777777778,
      "oracle_changed_fraction": 0.6041666666666666,
      "oracle_avg_edits": 0.8125,
      "oracle_in_candidate_set_fraction": 1.0,
      "iter_executor_accuracy": 0.2708333333333333,
      "iter_program_exact": 0.2708333333333333,
      "iter_state_prefix_fraction": 0.5989583333333334,
      "iter_changed_fraction": 0.0,
      "iter_avg_edits": 0.0,
      "iter_in_candidate_set_fraction": 1.0,
      "iter_steps_used_mean": 0.0,
      "iter_total_iter_edits_mean": 0.0,
      "iter_oracle_gap_recovered": 0.0
    },
    "val_len24_k1": {
      "n": 48.0,
      "k_steps": 1.0,
      "threshold": 0.8,
      "oracle_found_fraction": 0.875,
      "avg_candidates": 1299.0,
      "avg_positive_candidates": 1.125,
      "base_executor_accuracy": 0.2708333333333333,
      "base_program_exact": 0.2708333333333333,
      "base_state_prefix_fraction": 0.5989583333333334,
      "base_changed_fraction": 0.0,
      "base_avg_edits": 0.0,
      "base_in_candidate_set_fraction": 1.0,
      "oracle_executor_accuracy": 0.875,
      "oracle_program_exact": 0.875,
      "oracle_state_prefix_fraction": 0.9027777777777778,
      "oracle_changed_fraction": 0.6041666666666666,
      "oracle_avg_edits": 0.8125,
      "oracle_in_candidate_set_fraction": 1.0,
      "iter_executor_accuracy": 0.3333333333333333,
      "iter_program_exact": 0.3333333333333333,
      "iter_state_prefix_fraction": 0.6484375000000001,
      "iter_changed_fraction": 0.5416666666666666,
      "iter_avg_edits": 0.5416666666666666,
      "iter_in_candidate_set_fraction": 0.625,
      "iter_steps_used_mean": 0.5416666666666666,
      "iter_total_iter_edits_mean": 0.5416666666666666,
      "iter_oracle_gap_recovered": 0.10344827586206895
    },
    "val_len24_k2": {
      "n": 48.0,
      "k_steps": 2.0,
      "threshold": 0.8,
      "oracle_found_fraction": 0.875,
      "avg_candidates": 1299.0,
      "avg_positive_candidates": 1.125,
      "base_executor_accuracy": 0.2708333333333333,
      "base_program_exact": 0.2708333333333333,
      "base_state_prefix_fraction": 0.5989583333333334,
      "base_changed_fraction": 0.0,
      "base_avg_edits": 0.0,
      "base_in_candidate_set_fraction": 1.0,
      "oracle_executor_accuracy": 0.875,
      "oracle_program_exact": 0.875,
      "oracle_state_prefix_fraction": 0.9027777777777778,
      "oracle_changed_fraction": 0.6041666666666666,
      "oracle_avg_edits": 0.8125,
      "oracle_in_candidate_set_fraction": 1.0,
      "iter_executor_accuracy": 0.3125,
      "iter_program_exact": 0.3125,
      "iter_state_prefix_fraction": 0.631076388888889,
      "iter_changed_fraction": 0.5416666666666666,
      "iter_avg_edits": 0.5625,
      "iter_in_candidate_set_fraction": 0.625,
      "iter_steps_used_mean": 0.9375,
      "iter_total_iter_edits_mean": 0.9375,
      "iter_oracle_gap_recovered": 0.06896551724137934
    },
    "val_len24_k3": {
      "n": 48.0,
      "k_steps": 3.0,
      "threshold": 0.8,
      "oracle_found_fraction": 0.875,
      "avg_candidates": 1299.0,
      "avg_positive_candidates": 1.125,
      "base_executor_accuracy": 0.2708333333333333,
      "base_program_exact": 0.2708333333333333,
      "base_state_prefix_fraction": 0.5989583333333334,
      "base_changed_fraction": 0.0,
      "base_avg_edits": 0.0,
      "base_in_candidate_set_fraction": 1.0,
      "oracle_executor_accuracy": 0.875,
      "oracle_program_exact": 0.875,
      "oracle_state_prefix_fraction": 0.9027777777777778,
      "oracle_changed_fraction": 0.6041666666666666,
      "oracle_avg_edits": 0.8125,
      "oracle_in_candidate_set_fraction": 1.0,
      "iter_executor_accuracy": 0.3125,
      "iter_program_exact": 0.3125,
      "iter_state_prefix_fraction": 0.631076388888889,
      "iter_changed_fraction": 0.5416666666666666,
      "iter_avg_edits": 0.5625,
      "iter_in_candidate_set_fraction": 0.6041666666666666,
      "iter_steps_used_mean": 0.9791666666666666,
      "iter_total_iter_edits_mean": 0.9791666666666666,
      "iter_oracle_gap_recovered": 0.06896551724137934
    },
    "fresh_standard_len24_k0": {
      "n": 64.0,
      "k_steps": 0.0,
      "threshold": 0.8,
      "oracle_found_fraction": 0.90625,
      "avg_candidates": 1299.0,
      "avg_positive_candidates": 1.5,
      "base_executor_accuracy": 0.3125,
      "base_program_exact": 0.3125,
      "base_state_prefix_fraction": 0.5970052083333335,
      "base_changed_fraction": 0.0,
      "base_avg_edits": 0.0,
      "base_in_candidate_set_fraction": 1.0,
      "oracle_executor_accuracy": 0.90625,
      "oracle_program_exact": 0.90625,
      "oracle_state_prefix_fraction": 0.9264322916666665,
      "oracle_changed_fraction": 0.59375,
      "oracle_avg_edits": 0.84375,
      "oracle_in_candidate_set_fraction": 1.0,
      "iter_executor_accuracy": 0.3125,
      "iter_program_exact": 0.3125,
      "iter_state_prefix_fraction": 0.5970052083333335,
      "iter_changed_fraction": 0.0,
      "iter_avg_edits": 0.0,
      "iter_in_candidate_set_fraction": 1.0,
      "iter_steps_used_mean": 0.0,
      "iter_total_iter_edits_mean": 0.0,
      "iter_oracle_gap_recovered": 0.0
    },
    "fresh_standard_len24_k1": {
      "n": 64.0,
      "k_steps": 1.0,
      "threshold": 0.8,
      "oracle_found_fraction": 0.90625,
      "avg_candidates": 1299.0,
      "avg_positive_candidates": 1.5,
      "base_executor_accuracy": 0.3125,
      "base_program_exact": 0.3125,
      "base_state_prefix_fraction": 0.5970052083333335,
      "base_changed_fraction": 0.0,
      "base_avg_edits": 0.0,
      "base_in_candidate_set_fraction": 1.0,
      "oracle_executor_accuracy": 0.90625,
      "oracle_program_exact": 0.90625,
      "oracle_state_prefix_fraction": 0.9264322916666665,
      "oracle_changed_fraction": 0.59375,
      "oracle_avg_edits": 0.84375,
      "oracle_in_candidate_set_fraction": 1.0,
      "iter_executor_accuracy": 0.296875,
      "iter_program_exact": 0.296875,
      "iter_state_prefix_fraction": 0.6009114583333334,
      "iter_changed_fraction": 0.484375,
      "iter_avg_edits": 0.484375,
      "iter_in_candidate_set_fraction": 0.59375,
      "iter_steps_used_mean": 0.484375,
      "iter_total_iter_edits_mean": 0.484375,
      "iter_oracle_gap_recovered": -0.02631578947368421
    },
    "fresh_standard_len24_k2": {
      "n": 64.0,
      "k_steps": 2.0,
      "threshold": 0.8,
      "oracle_found_fraction": 0.90625,
      "avg_candidates": 1299.0,
      "avg_positive_candidates": 1.5,
      "base_executor_accuracy": 0.3125,
      "base_program_exact": 0.3125,
      "base_state_prefix_fraction": 0.5970052083333335,
      "base_changed_fraction": 0.0,
      "base_avg_edits": 0.0,
      "base_in_candidate_set_fraction": 1.0,
      "oracle_executor_accuracy": 0.90625,
      "oracle_program_exact": 0.90625,
      "oracle_state_prefix_fraction": 0.9264322916666665,
      "oracle_changed_fraction": 0.59375,
      "oracle_avg_edits": 0.84375,
      "oracle_in_candidate_set_fraction": 1.0,
      "iter_executor_accuracy": 0.296875,
      "iter_program_exact": 0.296875,
      "iter_state_prefix_fraction": 0.6009114583333334,
      "iter_changed_fraction": 0.484375,
      "iter_avg_edits": 0.484375,
      "iter_in_candidate_set_fraction": 0.625,
      "iter_steps_used_mean": 0.890625,
      "iter_total_iter_edits_mean": 0.890625,
      "iter_oracle_gap_recovered": -0.02631578947368421
    },
    "fresh_standard_len24_k3": {
      "n": 64.0,
      "k_steps": 3.0,
      "threshold": 0.8,
      "oracle_found_fraction": 0.90625,
      "avg_candidates": 1299.0,
      "avg_positive_candidates": 1.5,
      "base_executor_accuracy": 0.3125,
      "base_program_exact": 0.3125,
      "base_state_prefix_fraction": 0.5970052083333335,
      "base_changed_fraction": 0.0,
      "base_avg_edits": 0.0,
      "base_in_candidate_set_fraction": 1.0,
      "oracle_executor_accuracy": 0.90625,
      "oracle_program_exact": 0.90625,
      "oracle_state_prefix_fraction": 0.9264322916666665,
      "oracle_changed_fraction": 0.59375,
      "oracle_avg_edits": 0.84375,
      "oracle_in_candidate_set_fraction": 1.0,
      "iter_executor_accuracy": 0.296875,
      "iter_program_exact": 0.296875,
      "iter_state_prefix_fraction": 0.6009114583333334,
      "iter_changed_fraction": 0.484375,
      "iter_avg_edits": 0.484375,
      "iter_in_candidate_set_fraction": 0.59375,
      "iter_steps_used_mean": 0.9375,
      "iter_total_iter_edits_mean": 0.9375,
      "iter_oracle_gap_recovered": -0.02631578947368421
    },
    "fresh_paraphrase_len24_k0": {
      "n": 64.0,
      "k_steps": 0.0,
      "threshold": 0.8,
      "oracle_found_fraction": 0.890625,
      "avg_candidates": 1299.0,
      "avg_positive_candidates": 1.1875,
      "base_executor_accuracy": 0.234375,
      "base_program_exact": 0.234375,
      "base_state_prefix_fraction": 0.474609375,
      "base_changed_fraction": 0.0,
      "base_avg_edits": 0.0,
      "base_in_candidate_set_fraction": 1.0,
      "oracle_executor_accuracy": 0.890625,
      "oracle_program_exact": 0.890625,
      "oracle_state_prefix_fraction": 0.9082031249999999,
      "oracle_changed_fraction": 0.65625,
      "oracle_avg_edits": 0.9375,
      "oracle_in_candidate_set_fraction": 1.0,
      "iter_executor_accuracy": 0.234375,
      "iter_program_exact": 0.234375,
      "iter_state_prefix_fraction": 0.474609375,
      "iter_changed_fraction": 0.0,
      "iter_avg_edits": 0.0,
      "iter_in_candidate_set_fraction": 1.0,
      "iter_steps_used_mean": 0.0,
      "iter_total_iter_edits_mean": 0.0,
      "iter_oracle_gap_recovered": 0.0
    },
    "fresh_paraphrase_len24_k1": {
      "n": 64.0,
      "k_steps": 1.0,
      "threshold": 0.8,
      "oracle_found_fraction": 0.890625,
      "avg_candidates": 1299.0,
      "avg_positive_candidates": 1.1875,
      "base_executor_accuracy": 0.234375,
      "base_program_exact": 0.234375,
      "base_state_prefix_fraction": 0.474609375,
      "base_changed_fraction": 0.0,
      "base_avg_edits": 0.0,
      "base_in_candidate_set_fraction": 1.0,
      "oracle_executor_accuracy": 0.890625,
      "oracle_program_exact": 0.890625,
      "oracle_state_prefix_fraction": 0.9082031249999999,
      "oracle_changed_fraction": 0.65625,
      "oracle_avg_edits": 0.9375,
      "oracle_in_candidate_set_fraction": 1.0,
      "iter_executor_accuracy": 0.15625,
      "iter_program_exact": 0.15625,
      "iter_state_prefix_fraction": 0.43359374999999994,
      "iter_changed_fraction": 0.703125,
      "iter_avg_edits": 0.703125,
      "iter_in_candidate_set_fraction": 0.46875,
      "iter_steps_used_mean": 0.703125,
      "iter_total_iter_edits_mean": 0.703125,
      "iter_oracle_gap_recovered": -0.11904761904761904
    },
    "fresh_paraphrase_len24_k2": {
      "n": 64.0,
      "k_steps": 2.0,
      "threshold": 0.8,
      "oracle_found_fraction": 0.890625,
      "avg_candidates": 1299.0,
      "avg_positive_candidates": 1.1875,
      "base_executor_accuracy": 0.234375,
      "base_program_exact": 0.234375,
      "base_state_prefix_fraction": 0.474609375,
      "base_changed_fraction": 0.0,
      "base_avg_edits": 0.0,
      "base_in_candidate_set_fraction": 1.0,
      "oracle_executor_accuracy": 0.890625,
      "oracle_program_exact": 0.890625,
      "oracle_state_prefix_fraction": 0.9082031249999999,
      "oracle_changed_fraction": 0.65625,
      "oracle_avg_edits": 0.9375,
      "oracle_in_candidate_set_fraction": 1.0,
      "iter_executor_accuracy": 0.171875,
      "iter_program_exact": 0.171875,
      "iter_state_prefix_fraction": 0.4485677083333333,
      "iter_changed_fraction": 0.6875,
      "iter_avg_edits": 0.75,
      "iter_in_candidate_set_fraction": 0.453125,
      "iter_steps_used_mean": 1.3125,
      "iter_total_iter_edits_mean": 1.3125,
      "iter_oracle_gap_recovered": -0.09523809523809523
    },
    "fresh_paraphrase_len24_k3": {
      "n": 64.0,
      "k_steps": 3.0,
      "threshold": 0.8,
      "oracle_found_fraction": 0.890625,
      "avg_candidates": 1299.0,
      "avg_positive_candidates": 1.1875,
      "base_executor_accuracy": 0.234375,
      "base_program_exact": 0.234375,
      "base_state_prefix_fraction": 0.474609375,
      "base_changed_fraction": 0.0,
      "base_avg_edits": 0.0,
      "base_in_candidate_set_fraction": 1.0,
      "oracle_executor_accuracy": 0.890625,
      "oracle_program_exact": 0.890625,
      "oracle_state_prefix_fraction": 0.9082031249999999,
      "oracle_changed_fraction": 0.65625,
      "oracle_avg_edits": 0.9375,
      "oracle_in_candidate_set_fraction": 1.0,
      "iter_executor_accuracy": 0.171875,
      "iter_program_exact": 0.171875,
      "iter_state_prefix_fraction": 0.4485677083333333,
      "iter_changed_fraction": 0.703125,
      "iter_avg_edits": 0.765625,
      "iter_in_candidate_set_fraction": 0.421875,
      "iter_steps_used_mean": 1.375,
      "iter_total_iter_edits_mean": 1.375,
      "iter_oracle_gap_recovered": -0.09523809523809523
    },
    "fresh_paired_len24_k0": {
      "n": 128.0,
      "k_steps": 0.0,
      "threshold": 0.8,
      "oracle_found_fraction": 0.8828125,
      "avg_candidates": 1299.0,
      "avg_positive_candidates": 1.5078125,
      "base_executor_accuracy": 0.3203125,
      "base_program_exact": 0.3203125,
      "base_state_prefix_fraction": 0.5706380208333333,
      "base_changed_fraction": 0.0,
      "base_avg_edits": 0.0,
      "base_in_candidate_set_fraction": 1.0,
      "oracle_executor_accuracy": 0.8828125,
      "oracle_program_exact": 0.8828125,
      "oracle_state_prefix_fraction": 0.9055989583333333,
      "oracle_changed_fraction": 0.5625,
      "oracle_avg_edits": 0.7578125,
      "oracle_in_candidate_set_fraction": 1.0,
      "iter_executor_accuracy": 0.3203125,
      "iter_program_exact": 0.3203125,
      "iter_state_prefix_fraction": 0.5706380208333333,
      "iter_changed_fraction": 0.0,
      "iter_avg_edits": 0.0,
      "iter_in_candidate_set_fraction": 1.0,
      "iter_steps_used_mean": 0.0,
      "iter_total_iter_edits_mean": 0.0,
      "iter_oracle_gap_recovered": 0.0,
      "base_pair_answer_consistency": 0.703125,
      "base_pair_both_correct": 0.296875,
      "base_pair_program_consistency": 0.703125,
      "base_pair_state_consistency": 0.703125,
      "oracle_pair_answer_consistency": 0.90625,
      "oracle_pair_both_correct": 0.859375,
      "oracle_pair_program_consistency": 0.90625,
      "oracle_pair_state_consistency": 0.90625,
      "iter_pair_answer_consistency": 0.703125,
      "iter_pair_both_correct": 0.296875,
      "iter_pair_program_consistency": 0.703125,
      "iter_pair_state_consistency": 0.703125
    },
    "fresh_paired_len24_k1": {
      "n": 128.0,
      "k_steps": 1.0,
      "threshold": 0.8,
      "oracle_found_fraction": 0.8828125,
      "avg_candidates": 1299.0,
      "avg_positive_candidates": 1.5078125,
      "base_executor_accuracy": 0.3203125,
      "base_program_exact": 0.3203125,
      "base_state_prefix_fraction": 0.5706380208333333,
      "base_changed_fraction": 0.0,
      "base_avg_edits": 0.0,
      "base_in_candidate_set_fraction": 1.0,
      "oracle_executor_accuracy": 0.8828125,
      "oracle_program_exact": 0.8828125,
      "oracle_state_prefix_fraction": 0.9055989583333333,
      "oracle_changed_fraction": 0.5625,
      "oracle_avg_edits": 0.7578125,
      "oracle_in_candidate_set_fraction": 1.0,
      "iter_executor_accuracy": 0.2734375,
      "iter_program_exact": 0.2734375,
      "iter_state_prefix_fraction": 0.5345052083333333,
      "iter_changed_fraction": 0.609375,
      "iter_avg_edits": 0.609375,
      "iter_in_candidate_set_fraction": 0.5078125,
      "iter_steps_used_mean": 0.609375,
      "iter_total_iter_edits_mean": 0.609375,
      "iter_oracle_gap_recovered": -0.08333333333333333,
      "base_pair_answer_consistency": 0.703125,
      "base_pair_both_correct": 0.296875,
      "base_pair_program_consistency": 0.703125,
      "base_pair_state_consistency": 0.703125,
      "oracle_pair_answer_consistency": 0.90625,
      "oracle_pair_both_correct": 0.859375,
      "oracle_pair_program_consistency": 0.90625,
      "oracle_pair_state_consistency": 0.90625,
      "iter_pair_answer_consistency": 0.5625,
      "iter_pair_both_correct": 0.234375,
      "iter_pair_program_consistency": 0.546875,
      "iter_pair_state_consistency": 0.546875
    },
    "fresh_paired_len24_k2": {
      "n": 128.0,
      "k_steps": 2.0,
      "threshold": 0.8,
      "oracle_found_fraction": 0.8828125,
      "avg_candidates": 1299.0,
      "avg_positive_candidates": 1.5078125,
      "base_executor_accuracy": 0.3203125,
      "base_program_exact": 0.3203125,
      "base_state_prefix_fraction": 0.5706380208333333,
      "base_changed_fraction": 0.0,
      "base_avg_edits": 0.0,
      "base_in_candidate_set_fraction": 1.0,
      "oracle_executor_accuracy": 0.8828125,
      "oracle_program_exact": 0.8828125,
      "oracle_state_prefix_fraction": 0.9055989583333333,
      "oracle_changed_fraction": 0.5625,
      "oracle_avg_edits": 0.7578125,
      "oracle_in_candidate_set_fraction": 1.0,
      "iter_executor_accuracy": 0.2734375,
      "iter_program_exact": 0.2734375,
      "iter_state_prefix_fraction": 0.5345052083333333,
      "iter_changed_fraction": 0.59375,
      "iter_avg_edits": 0.625,
      "iter_in_candidate_set_fraction": 0.5,
      "iter_steps_used_mean": 1.1328125,
      "iter_total_iter_edits_mean": 1.1328125,
      "iter_oracle_gap_recovered": -0.08333333333333333,
      "base_pair_answer_consistency": 0.703125,
      "base_pair_both_correct": 0.296875,
      "base_pair_program_consistency": 0.703125,
      "base_pair_state_consistency": 0.703125,
      "oracle_pair_answer_consistency": 0.90625,
      "oracle_pair_both_correct": 0.859375,
      "oracle_pair_program_consistency": 0.90625,
      "oracle_pair_state_consistency": 0.90625,
      "iter_pair_answer_consistency": 0.546875,
      "iter_pair_both_correct": 0.234375,
      "iter_pair_program_consistency": 0.53125,
      "iter_pair_state_consistency": 0.53125
    },
    "fresh_paired_len24_k3": {
      "n": 128.0,
      "k_steps": 3.0,
      "threshold": 0.8,
      "oracle_found_fraction": 0.8828125,
      "avg_candidates": 1299.0,
      "avg_positive_candidates": 1.5078125,
      "base_executor_accuracy": 0.3203125,
      "base_program_exact": 0.3203125,
      "base_state_prefix_fraction": 0.5706380208333333,
      "base_changed_fraction": 0.0,
      "base_avg_edits": 0.0,
      "base_in_candidate_set_fraction": 1.0,
      "oracle_executor_accuracy": 0.8828125,
      "oracle_program_exact": 0.8828125,
      "oracle_state_prefix_fraction": 0.9055989583333333,
      "oracle_changed_fraction": 0.5625,
      "oracle_avg_edits": 0.7578125,
      "oracle_in_candidate_set_fraction": 1.0,
      "iter_executor_accuracy": 0.2734375,
      "iter_program_exact": 0.2734375,
      "iter_state_prefix_fraction": 0.5345052083333333,
      "iter_changed_fraction": 0.609375,
      "iter_avg_edits": 0.640625,
      "iter_in_candidate_set_fraction": 0.484375,
      "iter_steps_used_mean": 1.1953125,
      "iter_total_iter_edits_mean": 1.1953125,
      "iter_oracle_gap_recovered": -0.08333333333333333,
      "base_pair_answer_consistency": 0.703125,
      "base_pair_both_correct": 0.296875,
      "base_pair_program_consistency": 0.703125,
      "base_pair_state_consistency": 0.703125,
      "oracle_pair_answer_consistency": 0.90625,
      "oracle_pair_both_correct": 0.859375,
      "oracle_pair_program_consistency": 0.90625,
      "oracle_pair_state_consistency": 0.90625,
      "iter_pair_answer_consistency": 0.53125,
      "iter_pair_both_correct": 0.234375,
      "iter_pair_program_consistency": 0.515625,
      "iter_pair_state_consistency": 0.515625
    }
  },
  "train_log": [
    {
      "epoch": 1,
      "train_groups": 96,
      "train_cases": 384,
      "train_loss": 8.837961439043283,
      "train_gate_loss": 0.63203765230719,
      "train_value_loss": 8.205923768381277,
      "train_target_changed_slots": 1.1953125,
      "train_target_changed_fraction": 0.024394132653061156,
      "val_selected_threshold": 0.5,
      "val_selected_k": 1,
      "val_n": 48.0,
      "val_k_steps": 1.0,
      "val_threshold": 0.5,
      "val_oracle_found_fraction": 0.875,
      "val_avg_candidates": 1299.0,
      "val_avg_positive_candidates": 1.125,
      "val_base_executor_accuracy": 0.2708333333333333,
      "val_base_program_exact": 0.2708333333333333,
      "val_base_state_prefix_fraction": 0.5989583333333334,
      "val_base_changed_fraction": 0.0,
      "val_base_avg_edits": 0.0,
      "val_base_in_candidate_set_fraction": 1.0,
      "val_oracle_executor_accuracy": 0.875,
      "val_oracle_program_exact": 0.875,
      "val_oracle_state_prefix_fraction": 0.9027777777777778,
      "val_oracle_changed_fraction": 0.6041666666666666,
      "val_oracle_avg_edits": 0.8125,
      "val_oracle_in_candidate_set_fraction": 1.0,
      "val_iter_executor_accuracy": 0.3125,
      "val_iter_program_exact": 0.3125,
      "val_iter_state_prefix_fraction": 0.6206597222222222,
      "val_iter_changed_fraction": 0.375,
      "val_iter_avg_edits": 0.375,
      "val_iter_in_candidate_set_fraction": 0.7083333333333334,
      "val_iter_steps_used_mean": 0.375,
      "val_iter_total_iter_edits_mean": 0.375,
      "val_iter_oracle_gap_recovered": 0.06896551724137934
    },
    {
      "epoch": 2,
      "train_groups": 96,
      "train_cases": 384,
      "train_loss": 8.218887280672789,
      "train_gate_loss": 0.49259701105377945,
      "train_value_loss": 7.726290263235569,
      "train_target_changed_slots": 1.1640625,
      "train_target_changed_fraction": 0.02375637755102035,
      "val_selected_threshold": 0.8,
      "val_selected_k": 1,
      "val_n": 48.0,
      "val_k_steps": 1.0,
      "val_threshold": 0.8,
      "val_oracle_found_fraction": 0.875,
      "val_avg_candidates": 1299.0,
      "val_avg_positive_candidates": 1.125,
      "val_base_executor_accuracy": 0.2708333333333333,
      "val_base_program_exact": 0.2708333333333333,
      "val_base_state_prefix_fraction": 0.5989583333333334,
      "val_base_changed_fraction": 0.0,
      "val_base_avg_edits": 0.0,
      "val_base_in_candidate_set_fraction": 1.0,
      "val_oracle_executor_accuracy": 0.875,
      "val_oracle_program_exact": 0.875,
      "val_oracle_state_prefix_fraction": 0.9027777777777778,
      "val_oracle_changed_fraction": 0.6041666666666666,
      "val_oracle_avg_edits": 0.8125,
      "val_oracle_in_candidate_set_fraction": 1.0,
      "val_iter_executor_accuracy": 0.2708333333333333,
      "val_iter_program_exact": 0.2708333333333333,
      "val_iter_state_prefix_fraction": 0.5989583333333334,
      "val_iter_changed_fraction": 0.020833333333333332,
      "val_iter_avg_edits": 0.020833333333333332,
      "val_iter_in_candidate_set_fraction": 0.9791666666666666,
      "val_iter_steps_used_mean": 0.020833333333333332,
      "val_iter_total_iter_edits_mean": 0.020833333333333332,
      "val_iter_oracle_gap_recovered": 0.0
    },
    {
      "epoch": 3,
      "train_groups": 96,
      "train_cases": 384,
      "train_loss": 7.962621842821439,
      "train_gate_loss": 0.4639739429888626,
      "train_value_loss": 7.498647914578517,
      "train_target_changed_slots": 1.21875,
      "train_target_changed_fraction": 0.024872448979591774,
      "val_selected_threshold": 0.8,
      "val_selected_k": 1,
      "val_n": 48.0,
      "val_k_steps": 1.0,
      "val_threshold": 0.8,
      "val_oracle_found_fraction": 0.875,
      "val_avg_candidates": 1299.0,
      "val_avg_positive_candidates": 1.125,
      "val_base_executor_accuracy": 0.2708333333333333,
      "val_base_program_exact": 0.2708333333333333,
      "val_base_state_prefix_fraction": 0.5989583333333334,
      "val_base_changed_fraction": 0.0,
      "val_base_avg_edits": 0.0,
      "val_base_in_candidate_set_fraction": 1.0,
      "val_oracle_executor_accuracy": 0.875,
      "val_oracle_program_exact": 0.875,
      "val_oracle_state_prefix_fraction": 0.9027777777777778,
      "val_oracle_changed_fraction": 0.6041666666666666,
      "val_oracle_avg_edits": 0.8125,
      "val_oracle_in_candidate_set_fraction": 1.0,
      "val_iter_executor_accuracy": 0.3333333333333333,
      "val_iter_program_exact": 0.3333333333333333,
      "val_iter_state_prefix_fraction": 0.6484375000000001,
      "val_iter_changed_fraction": 0.5416666666666666,
      "val_iter_avg_edits": 0.5416666666666666,
      "val_iter_in_candidate_set_fraction": 0.625,
      "val_iter_steps_used_mean": 0.5416666666666666,
      "val_iter_total_iter_edits_mean": 0.5416666666666666,
      "val_iter_oracle_gap_recovered": 0.10344827586206895
    },
    {
      "epoch": 4,
      "train_groups": 96,
      "train_cases": 384,
      "train_loss": 7.63160773490866,
      "train_gate_loss": 0.412081481134616,
      "train_value_loss": 7.2195262387394905,
      "train_target_changed_slots": 1.1692708333333333,
      "train_target_changed_fraction": 0.023862670068027145,
      "val_selected_threshold": 0.8,
      "val_selected_k": 1,
      "val_n": 48.0,
      "val_k_steps": 1.0,
      "val_threshold": 0.8,
      "val_oracle_found_fraction": 0.875,
      "val_avg_candidates": 1299.0,
      "val_avg_positive_candidates": 1.125,
      "val_base_executor_accuracy": 0.2708333333333333,
      "val_base_program_exact": 0.2708333333333333,
      "val_base_state_prefix_fraction": 0.5989583333333334,
      "val_base_changed_fraction": 0.0,
      "val_base_avg_edits": 0.0,
      "val_base_in_candidate_set_fraction": 1.0,
      "val_oracle_executor_accuracy": 0.875,
      "val_oracle_program_exact": 0.875,
      "val_oracle_state_prefix_fraction": 0.9027777777777778,
      "val_oracle_changed_fraction": 0.6041666666666666,
      "val_oracle_avg_edits": 0.8125,
      "val_oracle_in_candidate_set_fraction": 1.0,
      "val_iter_executor_accuracy": 0.2916666666666667,
      "val_iter_program_exact": 0.2916666666666667,
      "val_iter_state_prefix_fraction": 0.6006944444444444,
      "val_iter_changed_fraction": 0.2916666666666667,
      "val_iter_avg_edits": 0.2916666666666667,
      "val_iter_in_candidate_set_fraction": 0.7916666666666666,
      "val_iter_steps_used_mean": 0.2916666666666667,
      "val_iter_total_iter_edits_mean": 0.2916666666666667,
      "val_iter_oracle_gap_recovered": 0.03448275862068971
    },
    {
      "epoch": 5,
      "train_groups": 96,
      "train_cases": 384,
      "train_loss": 7.482274889945984,
      "train_gate_loss": 0.3954188950689665,
      "train_value_loss": 7.086855987707774,
      "train_target_changed_slots": 1.1796875,
      "train_target_changed_fraction": 0.024075255102040758,
      "val_selected_threshold": 0.8,
      "val_selected_k": 1,
      "val_n": 48.0,
      "val_k_steps": 1.0,
      "val_threshold": 0.8,
      "val_oracle_found_fraction": 0.875,
      "val_avg_candidates": 1299.0,
      "val_avg_positive_candidates": 1.125,
      "val_base_executor_accuracy": 0.2708333333333333,
      "val_base_program_exact": 0.2708333333333333,
      "val_base_state_prefix_fraction": 0.5989583333333334,
      "val_base_changed_fraction": 0.0,
      "val_base_avg_edits": 0.0,
      "val_base_in_candidate_set_fraction": 1.0,
      "val_oracle_executor_accuracy": 0.875,
      "val_oracle_program_exact": 0.875,
      "val_oracle_state_prefix_fraction": 0.9027777777777778,
      "val_oracle_changed_fraction": 0.6041666666666666,
      "val_oracle_avg_edits": 0.8125,
      "val_oracle_in_candidate_set_fraction": 1.0,
      "val_iter_executor_accuracy": 0.22916666666666666,
      "val_iter_program_exact": 0.22916666666666666,
      "val_iter_state_prefix_fraction": 0.5460069444444445,
      "val_iter_changed_fraction": 0.7291666666666666,
      "val_iter_avg_edits": 0.7291666666666666,
      "val_iter_in_candidate_set_fraction": 0.4375,
      "val_iter_steps_used_mean": 0.7291666666666666,
      "val_iter_total_iter_edits_mean": 0.7291666666666666,
      "val_iter_oracle_gap_recovered": -0.06896551724137928
    },
    {
      "epoch": 6,
      "train_groups": 96,
      "train_cases": 384,
      "train_loss": 7.2910425166289015,
      "train_gate_loss": 0.36090626241154194,
      "train_value_loss": 6.930136264612277,
      "train_target_changed_slots": 1.1666666666666667,
      "train_target_changed_fraction": 0.02380952380952374,
      "val_selected_threshold": 0.7,
      "val_selected_k": 2,
      "val_n": 48.0,
      "val_k_steps": 2.0,
      "val_threshold": 0.7,
      "val_oracle_found_fraction": 0.875,
      "val_avg_candidates": 1299.0,
      "val_avg_positive_candidates": 1.125,
      "val_base_executor_accuracy": 0.2708333333333333,
      "val_base_program_exact": 0.2708333333333333,
      "val_base_state_prefix_fraction": 0.5989583333333334,
      "val_base_changed_fraction": 0.0,
      "val_base_avg_edits": 0.0,
      "val_base_in_candidate_set_fraction": 1.0,
      "val_oracle_executor_accuracy": 0.875,
      "val_oracle_program_exact": 0.875,
      "val_oracle_state_prefix_fraction": 0.9027777777777778,
      "val_oracle_changed_fraction": 0.6041666666666666,
      "val_oracle_avg_edits": 0.8125,
      "val_oracle_in_candidate_set_fraction": 1.0,
      "val_iter_executor_accuracy": 0.3125,
      "val_iter_program_exact": 0.3125,
      "val_iter_state_prefix_fraction": 0.6215277777777778,
      "val_iter_changed_fraction": 0.5416666666666666,
      "val_iter_avg_edits": 0.6666666666666666,
      "val_iter_in_candidate_set_fraction": 0.5625,
      "val_iter_steps_used_mean": 1.125,
      "val_iter_total_iter_edits_mean": 1.125,
      "val_iter_oracle_gap_recovered": 0.06896551724137934
    }
  ]
}