{
  "run": "smoke_process_dpo_20260624",
  "args": {
    "run_name": "smoke_process_dpo_20260624",
    "algorithm": "process_dpo",
    "model_name": "Qwen/Qwen3-4B",
    "seed": 431,
    "train_size": 4,
    "val_size": 2,
    "fresh_size": 2,
    "hard_size": 2,
    "max_arith_steps": 4,
    "max_prompt_len": 96,
    "state_tokens": 17,
    "dropout": 0.05,
    "lora_r": 8,
    "lora_alpha": 16,
    "lora_dropout": 0.05,
    "lr": 0.0001,
    "rl_lr": 5e-05,
    "bc_epochs": 1,
    "rl_epochs": 1,
    "rl_rounds": 1,
    "batch_size": 1,
    "grad_accum": 1,
    "teacher_max_steps": 8,
    "rl_rollout_steps": 3,
    "rollouts_per_prompt": 4,
    "temperature": 1.0,
    "stop_loss_weight": 4.0,
    "value_loss_weight": 0.5,
    "distance_loss_weight": 0.05,
    "rank_loss_weight": 1.0,
    "bc_echo_weight": 0.05,
    "echo_loss_weight": 0.05,
    "pg_loss_weight": 1.0,
    "rl_bc_loss_weight": 0.25,
    "rl_teacher_anchor_weight": 0.25,
    "false_stop_weight": 6.0,
    "preference_value_gate": 0,
    "preference_repair_max_edits": 4,
    "preference_repair_first_actions": 64,
    "preference_repair_second_actions": 48,
    "reward_correct": 1.0,
    "reward_exact_bonus": 0.25,
    "reward_false_stop_penalty": 0.25,
    "reward_invalid_penalty": 0.1,
    "reward_step_penalty": 0.01,
    "reward_shaping": 1,
    "reward_reachable_bonus": 0.05,
    "reward_destroy_reachability_penalty": 0.35,
    "reward_progress_weight": 0.08,
    "reward_regress_penalty": 0.05,
    "shape_repair_max_edits": 2,
    "shape_repair_first_actions": 48,
    "shape_repair_second_actions": 24,
    "shuffle_rollout_rewards": 0,
    "value_threshold": 0.55,
    "typed_action_mask": 1,
    "beam_width": 3,
    "beam_expand_actions": 4,
    "eval_native": 0,
    "native_max_new_tokens": 16,
    "k_values": [
      0,
      2
    ],
    "eval_modes": [
      "learned",
      "forced"
    ]
  },
  "metrics": [
    {
      "run": "smoke_process_dpo_20260624",
      "phase": "bc_policy",
      "split": "val_mixed",
      "mode": "blank",
      "k": 0,
      "n": 2,
      "accuracy": 0.0,
      "valid_rate": 1.0,
      "program_exact": 0.0,
      "stop_rate": 0.0,
      "false_stop_rate": 0.0,
      "mean_steps": 0.0,
      "blank_accuracy": 0.0,
      "oracle_accuracy": 0.0,
      "native_accuracy": 0.0
    },
    {
      "run": "smoke_process_dpo_20260624",
      "phase": "bc_policy",
      "split": "val_mixed",
      "mode": "oracle_teacher",
      "k": 0,
      "n": 2,
      "accuracy": 0.0,
      "valid_rate": 1.0,
      "program_exact": 0.0,
      "stop_rate": 0.0,
      "false_stop_rate": 0.0,
      "mean_steps": 0.0,
      "blank_accuracy": 0.0,
      "oracle_accuracy": 0.0,
      "native_accuracy": 0.0
    },
    {
      "run": "smoke_process_dpo_20260624",
      "phase": "bc_policy",
      "split": "val_mixed",
      "mode": "learned",
      "k": 0,
      "n": 2,
      "accuracy": 0.0,
      "valid_rate": 1.0,
      "program_exact": 0.0,
      "stop_rate": 0.0,
      "false_stop_rate": 0.0,
      "mean_steps": 0.0,
      "blank_accuracy": 0.0,
      "oracle_accuracy": 0.0,
      "native_accuracy": 0.0
    },
    {
      "run": "smoke_process_dpo_20260624",
      "phase": "bc_policy",
      "split": "val_mixed",
      "mode": "forced",
      "k": 0,
      "n": 2,
      "accuracy": 0.0,
      "valid_rate": 1.0,
      "program_exact": 0.0,
      "stop_rate": 0.0,
      "false_stop_rate": 0.0,
      "mean_steps": 0.0,
      "blank_accuracy": 0.0,
      "oracle_accuracy": 0.0,
      "native_accuracy": 0.0
    },
    {
      "run": "smoke_process_dpo_20260624",
      "phase": "bc_policy",
      "split": "val_mixed",
      "mode": "oracle_teacher",
      "k": 2,
      "n": 2,
      "accuracy": 0.0,
      "valid_rate": 0.0,
      "program_exact": 0.0,
      "stop_rate": 0.0,
      "false_stop_rate": 0.0,
      "mean_steps": 2.0,
      "blank_accuracy": 0.0,
      "oracle_accuracy": 0.0,
      "native_accuracy": 0.0
    },
    {
      "run": "smoke_process_dpo_20260624",
      "phase": "bc_policy",
      "split": "val_mixed",
      "mode": "learned",
      "k": 2,
      "n": 2,
      "accuracy": 0.0,
      "valid_rate": 1.0,
      "program_exact": 0.0,
      "stop_rate": 1.0,
      "false_stop_rate": 1.0,
      "mean_steps": 0.0,
      "blank_accuracy": 0.0,
      "oracle_accuracy": 0.0,
      "native_accuracy": 0.0
    },
    {
      "run": "smoke_process_dpo_20260624",
      "phase": "bc_policy",
      "split": "val_mixed",
      "mode": "forced",
      "k": 2,
      "n": 2,
      "accuracy": 0.5,
      "valid_rate": 1.0,
      "program_exact": 0.0,
      "stop_rate": 0.0,
      "false_stop_rate": 0.0,
      "mean_steps": 2.0,
      "blank_accuracy": 0.0,
      "oracle_accuracy": 0.0,
      "native_accuracy": 0.0
    },
    {
      "run": "smoke_process_dpo_20260624",
      "phase": "bc_policy",
      "split": "fresh_standard",
      "mode": "blank",
      "k": 0,
      "n": 2,
      "accuracy": 0.0,
      "valid_rate": 1.0,
      "program_exact": 0.0,
      "stop_rate": 0.0,
      "false_stop_rate": 0.0,
      "mean_steps": 0.0,
      "blank_accuracy": 0.0,
      "oracle_accuracy": 0.0,
      "native_accuracy": 0.0
    },
    {
      "run": "smoke_process_dpo_20260624",
      "phase": "bc_policy",
      "split": "fresh_standard",
      "mode": "oracle_teacher",
      "k": 0,
      "n": 2,
      "accuracy": 0.0,
      "valid_rate": 1.0,
      "program_exact": 0.0,
      "stop_rate": 0.0,
      "false_stop_rate": 0.0,
      "mean_steps": 0.0,
      "blank_accuracy": 0.0,
      "oracle_accuracy": 0.0,
      "native_accuracy": 0.0
    },
    {
      "run": "smoke_process_dpo_20260624",
      "phase": "bc_policy",
      "split": "fresh_standard",
      "mode": "learned",
      "k": 0,
      "n": 2,
      "accuracy": 0.0,
      "valid_rate": 1.0,
      "program_exact": 0.0,
      "stop_rate": 0.0,
      "false_stop_rate": 0.0,
      "mean_steps": 0.0,
      "blank_accuracy": 0.0,
      "oracle_accuracy": 0.0,
      "native_accuracy": 0.0
    },
    {
      "run": "smoke_process_dpo_20260624",
      "phase": "bc_policy",
      "split": "fresh_standard",
      "mode": "forced",
      "k": 0,
      "n": 2,
      "accuracy": 0.0,
      "valid_rate": 1.0,
      "program_exact": 0.0,
      "stop_rate": 0.0,
      "false_stop_rate": 0.0,
      "mean_steps": 0.0,
      "blank_accuracy": 0.0,
      "oracle_accuracy": 0.0,
      "native_accuracy": 0.0
    },
    {
      "run": "smoke_process_dpo_20260624",
      "phase": "bc_policy",
      "split": "fresh_standard",
      "mode": "oracle_teacher",
      "k": 2,
      "n": 2,
      "accuracy": 0.0,
      "valid_rate": 0.5,
      "program_exact": 0.0,
      "stop_rate": 0.0,
      "false_stop_rate": 0.0,
      "mean_steps": 2.0,
      "blank_accuracy": 0.0,
      "oracle_accuracy": 0.0,
      "native_accuracy": 0.0
    },
    {
      "run": "smoke_process_dpo_20260624",
      "phase": "bc_policy",
      "split": "fresh_standard",
      "mode": "learned",
      "k": 2,
      "n": 2,
      "accuracy": 0.0,
      "valid_rate": 1.0,
      "program_exact": 0.0,
      "stop_rate": 1.0,
      "false_stop_rate": 1.0,
      "mean_steps": 0.0,
      "blank_accuracy": 0.0,
      "oracle_accuracy": 0.0,
      "native_accuracy": 0.0
    },
    {
      "run": "smoke_process_dpo_20260624",
      "phase": "bc_policy",
      "split": "fresh_standard",
      "mode": "forced",
      "k": 2,
      "n": 2,
      "accuracy": 0.0,
      "valid_rate": 1.0,
      "program_exact": 0.0,
      "stop_rate": 0.0,
      "false_stop_rate": 0.0,
      "mean_steps": 2.0,
      "blank_accuracy": 0.0,
      "oracle_accuracy": 0.0,
      "native_accuracy": 0.0
    },
    {
      "run": "smoke_process_dpo_20260624",
      "phase": "bc_policy",
      "split": "fresh_paraphrase",
      "mode": "blank",
      "k": 0,
      "n": 2,
      "accuracy": 0.0,
      "valid_rate": 1.0,
      "program_exact": 0.0,
      "stop_rate": 0.0,
      "false_stop_rate": 0.0,
      "mean_steps": 0.0,
      "blank_accuracy": 0.0,
      "oracle_accuracy": 0.0,
      "native_accuracy": 0.0
    },
    {
      "run": "smoke_process_dpo_20260624",
      "phase": "bc_policy",
      "split": "fresh_paraphrase",
      "mode": "oracle_teacher",
      "k": 0,
      "n": 2,
      "accuracy": 0.0,
      "valid_rate": 1.0,
      "program_exact": 0.0,
      "stop_rate": 0.0,
      "false_stop_rate": 0.0,
      "mean_steps": 0.0,
      "blank_accuracy": 0.0,
      "oracle_accuracy": 0.0,
      "native_accuracy": 0.0
    },
    {
      "run": "smoke_process_dpo_20260624",
      "phase": "bc_policy",
      "split": "fresh_paraphrase",
      "mode": "learned",
      "k": 0,
      "n": 2,
      "accuracy": 0.0,
      "valid_rate": 1.0,
      "program_exact": 0.0,
      "stop_rate": 0.0,
      "false_stop_rate": 0.0,
      "mean_steps": 0.0,
      "blank_accuracy": 0.0,
      "oracle_accuracy": 0.0,
      "native_accuracy": 0.0
    },
    {
      "run": "smoke_process_dpo_20260624",
      "phase": "bc_policy",
      "split": "fresh_paraphrase",
      "mode": "forced",
      "k": 0,
      "n": 2,
      "accuracy": 0.0,
      "valid_rate": 1.0,
      "program_exact": 0.0,
      "stop_rate": 0.0,
      "false_stop_rate": 0.0,
      "mean_steps": 0.0,
      "blank_accuracy": 0.0,
      "oracle_accuracy": 0.0,
      "native_accuracy": 0.0
    },
    {
      "run": "smoke_process_dpo_20260624",
      "phase": "bc_policy",
      "split": "fresh_paraphrase",
      "mode": "oracle_teacher",
      "k": 2,
      "n": 2,
      "accuracy": 0.0,
      "valid_rate": 0.0,
      "program_exact": 0.0,
      "stop_rate": 0.0,
      "false_stop_rate": 0.0,
      "mean_steps": 2.0,
      "blank_accuracy": 0.0,
      "oracle_accuracy": 0.0,
      "native_accuracy": 0.0
    },
    {
      "run": "smoke_process_dpo_20260624",
      "phase": "bc_policy",
      "split": "fresh_paraphrase",
      "mode": "learned",
      "k": 2,
      "n": 2,
      "accuracy": 0.0,
      "valid_rate": 1.0,
      "program_exact": 0.0,
      "stop_rate": 1.0,
      "false_stop_rate": 1.0,
      "mean_steps": 0.0,
      "blank_accuracy": 0.0,
      "oracle_accuracy": 0.0,
      "native_accuracy": 0.0
    },
    {
      "run": "smoke_process_dpo_20260624",
      "phase": "bc_policy",
      "split": "fresh_paraphrase",
      "mode": "forced",
      "k": 2,
      "n": 2,
      "accuracy": 0.0,
      "valid_rate": 1.0,
      "program_exact": 0.0,
      "stop_rate": 0.0,
      "false_stop_rate": 0.0,
      "mean_steps": 2.0,
      "blank_accuracy": 0.0,
      "oracle_accuracy": 0.0,
      "native_accuracy": 0.0
    },
    {
      "run": "smoke_process_dpo_20260624",
      "phase": "bc_policy",
      "split": "fresh_paired",
      "mode": "blank",
      "k": 0,
      "n": 2,
      "accuracy": 0.0,
      "valid_rate": 1.0,
      "program_exact": 0.0,
      "stop_rate": 0.0,
      "false_stop_rate": 0.0,
      "mean_steps": 0.0,
      "blank_accuracy": 0.0,
      "oracle_accuracy": 0.0,
      "native_accuracy": 0.0
    },
    {
      "run": "smoke_process_dpo_20260624",
      "phase": "bc_policy",
      "split": "fresh_paired",
      "mode": "oracle_teacher",
      "k": 0,
      "n": 2,
      "accuracy": 0.0,
      "valid_rate": 1.0,
      "program_exact": 0.0,
      "stop_rate": 0.0,
      "false_stop_rate": 0.0,
      "mean_steps": 0.0,
      "blank_accuracy": 0.0,
      "oracle_accuracy": 0.0,
      "native_accuracy": 0.0
    },
    {
      "run": "smoke_process_dpo_20260624",
      "phase": "bc_policy",
      "split": "fresh_paired",
      "mode": "learned",
      "k": 0,
      "n": 2,
      "accuracy": 0.0,
      "valid_rate": 1.0,
      "program_exact": 0.0,
      "stop_rate": 0.0,
      "false_stop_rate": 0.0,
      "mean_steps": 0.0,
      "blank_accuracy": 0.0,
      "oracle_accuracy": 0.0,
      "native_accuracy": 0.0
    },
    {
      "run": "smoke_process_dpo_20260624",
      "phase": "bc_policy",
      "split": "fresh_paired",
      "mode": "forced",
      "k": 0,
      "n": 2,
      "accuracy": 0.0,
      "valid_rate": 1.0,
      "program_exact": 0.0,
      "stop_rate": 0.0,
      "false_stop_rate": 0.0,
      "mean_steps": 0.0,
      "blank_accuracy": 0.0,
      "oracle_accuracy": 0.0,
      "native_accuracy": 0.0
    },
    {
      "run": "smoke_process_dpo_20260624",
      "phase": "bc_policy",
      "split": "fresh_paired",
      "mode": "oracle_teacher",
      "k": 2,
      "n": 2,
      "accuracy": 0.0,
      "valid_rate": 1.0,
      "program_exact": 0.0,
      "stop_rate": 0.0,
      "false_stop_rate": 0.0,
      "mean_steps": 2.0,
      "blank_accuracy": 0.0,
      "oracle_accuracy": 0.0,
      "native_accuracy": 0.0
    },
    {
      "run": "smoke_process_dpo_20260624",
      "phase": "bc_policy",
      "split": "fresh_paired",
      "mode": "learned",
      "k": 2,
      "n": 2,
      "accuracy": 0.0,
      "valid_rate": 1.0,
      "program_exact": 0.0,
      "stop_rate": 1.0,
      "false_stop_rate": 1.0,
      "mean_steps": 0.0,
      "blank_accuracy": 0.0,
      "oracle_accuracy": 0.0,
      "native_accuracy": 0.0
    },
    {
      "run": "smoke_process_dpo_20260624",
      "phase": "bc_policy",
      "split": "fresh_paired",
      "mode": "forced",
      "k": 2,
      "n": 2,
      "accuracy": 0.0,
      "valid_rate": 1.0,
      "program_exact": 0.0,
      "stop_rate": 0.0,
      "false_stop_rate": 0.0,
      "mean_steps": 2.0,
      "blank_accuracy": 0.0,
      "oracle_accuracy": 0.0,
      "native_accuracy": 0.0
    },
    {
      "run": "smoke_process_dpo_20260624",
      "phase": "bc_policy",
      "split": "hard_composition",
      "mode": "blank",
      "k": 0,
      "n": 2,
      "accuracy": 0.5,
      "valid_rate": 1.0,
      "program_exact": 0.0,
      "stop_rate": 0.0,
      "false_stop_rate": 0.0,
      "mean_steps": 0.0,
      "blank_accuracy": 0.5,
      "oracle_accuracy": 0.5,
      "native_accuracy": 0.0
    },
    {
      "run": "smoke_process_dpo_20260624",
      "phase": "bc_policy",
      "split": "hard_composition",
      "mode": "oracle_teacher",
      "k": 0,
      "n": 2,
      "accuracy": 0.5,
      "valid_rate": 1.0,
      "program_exact": 0.0,
      "stop_rate": 0.0,
      "false_stop_rate": 0.0,
      "mean_steps": 0.0,
      "blank_accuracy": 0.5,
      "oracle_accuracy": 0.5,
      "native_accuracy": 0.0
    },
    {
      "run": "smoke_process_dpo_20260624",
      "phase": "bc_policy",
      "split": "hard_composition",
      "mode": "learned",
      "k": 0,
      "n": 2,
      "accuracy": 0.5,
      "valid_rate": 1.0,
      "program_exact": 0.0,
      "stop_rate": 0.0,
      "false_stop_rate": 0.0,
      "mean_steps": 0.0,
      "blank_accuracy": 0.5,
      "oracle_accuracy": 0.5,
      "native_accuracy": 0.0
    },
    {
      "run": "smoke_process_dpo_20260624",
      "phase": "bc_policy",
      "split": "hard_composition",
      "mode": "forced",
      "k": 0,
      "n": 2,
      "accuracy": 0.5,
      "valid_rate": 1.0,
      "program_exact": 0.0,
      "stop_rate": 0.0,
      "false_stop_rate": 0.0,
      "mean_steps": 0.0,
      "blank_accuracy": 0.5,
      "oracle_accuracy": 0.5,
      "native_accuracy": 0.0
    },
    {
      "run": "smoke_process_dpo_20260624",
      "phase": "bc_policy",
      "split": "hard_composition",
      "mode": "oracle_teacher",
      "k": 2,
      "n": 2,
      "accuracy": 0.5,
      "valid_rate": 0.5,
      "program_exact": 0.0,
      "stop_rate": 0.5,
      "false_stop_rate": 0.0,
      "mean_steps": 1.0,
      "blank_accuracy": 0.5,
      "oracle_accuracy": 0.5,
      "native_accuracy": 0.0
    },
    {
      "run": "smoke_process_dpo_20260624",
      "phase": "bc_policy",
      "split": "hard_composition",
      "mode": "learned",
      "k": 2,
      "n": 2,
      "accuracy": 0.5,
      "valid_rate": 1.0,
      "program_exact": 0.0,
      "stop_rate": 1.0,
      "false_stop_rate": 0.5,
      "mean_steps": 0.0,
      "blank_accuracy": 0.5,
      "oracle_accuracy": 0.5,
      "native_accuracy": 0.0
    },
    {
      "run": "smoke_process_dpo_20260624",
      "phase": "bc_policy",
      "split": "hard_composition",
      "mode": "forced",
      "k": 2,
      "n": 2,
      "accuracy": 0.0,
      "valid_rate": 1.0,
      "program_exact": 0.0,
      "stop_rate": 0.0,
      "false_stop_rate": 0.0,
      "mean_steps": 2.0,
      "blank_accuracy": 0.5,
      "oracle_accuracy": 0.5,
      "native_accuracy": 0.0
    },
    {
      "run": "smoke_process_dpo_20260624",
      "phase": "process_dpo_r1_policy",
      "split": "val_mixed",
      "mode": "blank",
      "k": 0,
      "n": 2,
      "accuracy": 0.0,
      "valid_rate": 1.0,
      "program_exact": 0.0,
      "stop_rate": 0.0,
      "false_stop_rate": 0.0,
      "mean_steps": 0.0,
      "blank_accuracy": 0.0,
      "oracle_accuracy": 0.0,
      "native_accuracy": 0.0
    },
    {
      "run": "smoke_process_dpo_20260624",
      "phase": "process_dpo_r1_policy",
      "split": "val_mixed",
      "mode": "oracle_teacher",
      "k": 0,
      "n": 2,
      "accuracy": 0.0,
      "valid_rate": 1.0,
      "program_exact": 0.0,
      "stop_rate": 0.0,
      "false_stop_rate": 0.0,
      "mean_steps": 0.0,
      "blank_accuracy": 0.0,
      "oracle_accuracy": 0.0,
      "native_accuracy": 0.0
    },
    {
      "run": "smoke_process_dpo_20260624",
      "phase": "process_dpo_r1_policy",
      "split": "val_mixed",
      "mode": "learned",
      "k": 0,
      "n": 2,
      "accuracy": 0.0,
      "valid_rate": 1.0,
      "program_exact": 0.0,
      "stop_rate": 0.0,
      "false_stop_rate": 0.0,
      "mean_steps": 0.0,
      "blank_accuracy": 0.0,
      "oracle_accuracy": 0.0,
      "native_accuracy": 0.0
    },
    {
      "run": "smoke_process_dpo_20260624",
      "phase": "process_dpo_r1_policy",
      "split": "val_mixed",
      "mode": "forced",
      "k": 0,
      "n": 2,
      "accuracy": 0.0,
      "valid_rate": 1.0,
      "program_exact": 0.0,
      "stop_rate": 0.0,
      "false_stop_rate": 0.0,
      "mean_steps": 0.0,
      "blank_accuracy": 0.0,
      "oracle_accuracy": 0.0,
      "native_accuracy": 0.0
    },
    {
      "run": "smoke_process_dpo_20260624",
      "phase": "process_dpo_r1_policy",
      "split": "val_mixed",
      "mode": "oracle_teacher",
      "k": 2,
      "n": 2,
      "accuracy": 0.0,
      "valid_rate": 0.0,
      "program_exact": 0.0,
      "stop_rate": 0.0,
      "false_stop_rate": 0.0,
      "mean_steps": 2.0,
      "blank_accuracy": 0.0,
      "oracle_accuracy": 0.0,
      "native_accuracy": 0.0
    },
    {
      "run": "smoke_process_dpo_20260624",
      "phase": "process_dpo_r1_policy",
      "split": "val_mixed",
      "mode": "learned",
      "k": 2,
      "n": 2,
      "accuracy": 0.0,
      "valid_rate": 0.0,
      "program_exact": 0.0,
      "stop_rate": 0.0,
      "false_stop_rate": 0.0,
      "mean_steps": 2.0,
      "blank_accuracy": 0.0,
      "oracle_accuracy": 0.0,
      "native_accuracy": 0.0
    },
    {
      "run": "smoke_process_dpo_20260624",
      "phase": "process_dpo_r1_policy",
      "split": "val_mixed",
      "mode": "forced",
      "k": 2,
      "n": 2,
      "accuracy": 0.0,
      "valid_rate": 0.0,
      "program_exact": 0.0,
      "stop_rate": 0.0,
      "false_stop_rate": 0.0,
      "mean_steps": 2.0,
      "blank_accuracy": 0.0,
      "oracle_accuracy": 0.0,
      "native_accuracy": 0.0
    },
    {
      "run": "smoke_process_dpo_20260624",
      "phase": "process_dpo_r1_policy",
      "split": "fresh_standard",
      "mode": "blank",
      "k": 0,
      "n": 2,
      "accuracy": 0.0,
      "valid_rate": 1.0,
      "program_exact": 0.0,
      "stop_rate": 0.0,
      "false_stop_rate": 0.0,
      "mean_steps": 0.0,
      "blank_accuracy": 0.0,
      "oracle_accuracy": 0.0,
      "native_accuracy": 0.0
    },
    {
      "run": "smoke_process_dpo_20260624",
      "phase": "process_dpo_r1_policy",
      "split": "fresh_standard",
      "mode": "oracle_teacher",
      "k": 0,
      "n": 2,
      "accuracy": 0.0,
      "valid_rate": 1.0,
      "program_exact": 0.0,
      "stop_rate": 0.0,
      "false_stop_rate": 0.0,
      "mean_steps": 0.0,
      "blank_accuracy": 0.0,
      "oracle_accuracy": 0.0,
      "native_accuracy": 0.0
    },
    {
      "run": "smoke_process_dpo_20260624",
      "phase": "process_dpo_r1_policy",
      "split": "fresh_standard",
      "mode": "learned",
      "k": 0,
      "n": 2,
      "accuracy": 0.0,
      "valid_rate": 1.0,
      "program_exact": 0.0,
      "stop_rate": 0.0,
      "false_stop_rate": 0.0,
      "mean_steps": 0.0,
      "blank_accuracy": 0.0,
      "oracle_accuracy": 0.0,
      "native_accuracy": 0.0
    },
    {
      "run": "smoke_process_dpo_20260624",
      "phase": "process_dpo_r1_policy",
      "split": "fresh_standard",
      "mode": "forced",
      "k": 0,
      "n": 2,
      "accuracy": 0.0,
      "valid_rate": 1.0,
      "program_exact": 0.0,
      "stop_rate": 0.0,
      "false_stop_rate": 0.0,
      "mean_steps": 0.0,
      "blank_accuracy": 0.0,
      "oracle_accuracy": 0.0,
      "native_accuracy": 0.0
    },
    {
      "run": "smoke_process_dpo_20260624",
      "phase": "process_dpo_r1_policy",
      "split": "fresh_standard",
      "mode": "oracle_teacher",
      "k": 2,
      "n": 2,
      "accuracy": 0.0,
      "valid_rate": 0.5,
      "program_exact": 0.0,
      "stop_rate": 0.0,
      "false_stop_rate": 0.0,
      "mean_steps": 2.0,
      "blank_accuracy": 0.0,
      "oracle_accuracy": 0.0,
      "native_accuracy": 0.0
    },
    {
      "run": "smoke_process_dpo_20260624",
      "phase": "process_dpo_r1_policy",
      "split": "fresh_standard",
      "mode": "learned",
      "k": 2,
      "n": 2,
      "accuracy": 0.0,
      "valid_rate": 0.0,
      "program_exact": 0.0,
      "stop_rate": 0.0,
      "false_stop_rate": 0.0,
      "mean_steps": 2.0,
      "blank_accuracy": 0.0,
      "oracle_accuracy": 0.0,
      "native_accuracy": 0.0
    },
    {
      "run": "smoke_process_dpo_20260624",
      "phase": "process_dpo_r1_policy",
      "split": "fresh_standard",
      "mode": "forced",
      "k": 2,
      "n": 2,
      "accuracy": 0.0,
      "valid_rate": 0.0,
      "program_exact": 0.0,
      "stop_rate": 0.0,
      "false_stop_rate": 0.0,
      "mean_steps": 2.0,
      "blank_accuracy": 0.0,
      "oracle_accuracy": 0.0,
      "native_accuracy": 0.0
    },
    {
      "run": "smoke_process_dpo_20260624",
      "phase": "process_dpo_r1_policy",
      "split": "fresh_paraphrase",
      "mode": "blank",
      "k": 0,
      "n": 2,
      "accuracy": 0.0,
      "valid_rate": 1.0,
      "program_exact": 0.0,
      "stop_rate": 0.0,
      "false_stop_rate": 0.0,
      "mean_steps": 0.0,
      "blank_accuracy": 0.0,
      "oracle_accuracy": 0.0,
      "native_accuracy": 0.0
    },
    {
      "run": "smoke_process_dpo_20260624",
      "phase": "process_dpo_r1_policy",
      "split": "fresh_paraphrase",
      "mode": "oracle_teacher",
      "k": 0,
      "n": 2,
      "accuracy": 0.0,
      "valid_rate": 1.0,
      "program_exact": 0.0,
      "stop_rate": 0.0,
      "false_stop_rate": 0.0,
      "mean_steps": 0.0,
      "blank_accuracy": 0.0,
      "oracle_accuracy": 0.0,
      "native_accuracy": 0.0
    },
    {
      "run": "smoke_process_dpo_20260624",
      "phase": "process_dpo_r1_policy",
      "split": "fresh_paraphrase",
      "mode": "learned",
      "k": 0,
      "n": 2,
      "accuracy": 0.0,
      "valid_rate": 1.0,
      "program_exact": 0.0,
      "stop_rate": 0.0,
      "false_stop_rate": 0.0,
      "mean_steps": 0.0,
      "blank_accuracy": 0.0,
      "oracle_accuracy": 0.0,
      "native_accuracy": 0.0
    },
    {
      "run": "smoke_process_dpo_20260624",
      "phase": "process_dpo_r1_policy",
      "split": "fresh_paraphrase",
      "mode": "forced",
      "k": 0,
      "n": 2,
      "accuracy": 0.0,
      "valid_rate": 1.0,
      "program_exact": 0.0,
      "stop_rate": 0.0,
      "false_stop_rate": 0.0,
      "mean_steps": 0.0,
      "blank_accuracy": 0.0,
      "oracle_accuracy": 0.0,
      "native_accuracy": 0.0
    },
    {
      "run": "smoke_process_dpo_20260624",
      "phase": "process_dpo_r1_policy",
      "split": "fresh_paraphrase",
      "mode": "oracle_teacher",
      "k": 2,
      "n": 2,
      "accuracy": 0.0,
      "valid_rate": 0.0,
      "program_exact": 0.0,
      "stop_rate": 0.0,
      "false_stop_rate": 0.0,
      "mean_steps": 2.0,
      "blank_accuracy": 0.0,
      "oracle_accuracy": 0.0,
      "native_accuracy": 0.0
    },
    {
      "run": "smoke_process_dpo_20260624",
      "phase": "process_dpo_r1_policy",
      "split": "fresh_paraphrase",
      "mode": "learned",
      "k": 2,
      "n": 2,
      "accuracy": 0.0,
      "valid_rate": 0.0,
      "program_exact": 0.0,
      "stop_rate": 0.0,
      "false_stop_rate": 0.0,
      "mean_steps": 2.0,
      "blank_accuracy": 0.0,
      "oracle_accuracy": 0.0,
      "native_accuracy": 0.0
    },
    {
      "run": "smoke_process_dpo_20260624",
      "phase": "process_dpo_r1_policy",
      "split": "fresh_paraphrase",
      "mode": "forced",
      "k": 2,
      "n": 2,
      "accuracy": 0.0,
      "valid_rate": 0.0,
      "program_exact": 0.0,
      "stop_rate": 0.0,
      "false_stop_rate": 0.0,
      "mean_steps": 2.0,
      "blank_accuracy": 0.0,
      "oracle_accuracy": 0.0,
      "native_accuracy": 0.0
    },
    {
      "run": "smoke_process_dpo_20260624",
      "phase": "process_dpo_r1_policy",
      "split": "fresh_paired",
      "mode": "blank",
      "k": 0,
      "n": 2,
      "accuracy": 0.0,
      "valid_rate": 1.0,
      "program_exact": 0.0,
      "stop_rate": 0.0,
      "false_stop_rate": 0.0,
      "mean_steps": 0.0,
      "blank_accuracy": 0.0,
      "oracle_accuracy": 0.0,
      "native_accuracy": 0.0
    },
    {
      "run": "smoke_process_dpo_20260624",
      "phase": "process_dpo_r1_policy",
      "split": "fresh_paired",
      "mode": "oracle_teacher",
      "k": 0,
      "n": 2,
      "accuracy": 0.0,
      "valid_rate": 1.0,
      "program_exact": 0.0,
      "stop_rate": 0.0,
      "false_stop_rate": 0.0,
      "mean_steps": 0.0,
      "blank_accuracy": 0.0,
      "oracle_accuracy": 0.0,
      "native_accuracy": 0.0
    },
    {
      "run": "smoke_process_dpo_20260624",
      "phase": "process_dpo_r1_policy",
      "split": "fresh_paired",
      "mode": "learned",
      "k": 0,
      "n": 2,
      "accuracy": 0.0,
      "valid_rate": 1.0,
      "program_exact": 0.0,
      "stop_rate": 0.0,
      "false_stop_rate": 0.0,
      "mean_steps": 0.0,
      "blank_accuracy": 0.0,
      "oracle_accuracy": 0.0,
      "native_accuracy": 0.0
    },
    {
      "run": "smoke_process_dpo_20260624",
      "phase": "process_dpo_r1_policy",
      "split": "fresh_paired",
      "mode": "forced",
      "k": 0,
      "n": 2,
      "accuracy": 0.0,
      "valid_rate": 1.0,
      "program_exact": 0.0,
      "stop_rate": 0.0,
      "false_stop_rate": 0.0,
      "mean_steps": 0.0,
      "blank_accuracy": 0.0,
      "oracle_accuracy": 0.0,
      "native_accuracy": 0.0
    },
    {
      "run": "smoke_process_dpo_20260624",
      "phase": "process_dpo_r1_policy",
      "split": "fresh_paired",
      "mode": "oracle_teacher",
      "k": 2,
      "n": 2,
      "accuracy": 0.0,
      "valid_rate": 1.0,
      "program_exact": 0.0,
      "stop_rate": 0.0,
      "false_stop_rate": 0.0,
      "mean_steps": 2.0,
      "blank_accuracy": 0.0,
      "oracle_accuracy": 0.0,
      "native_accuracy": 0.0
    },
    {
      "run": "smoke_process_dpo_20260624",
      "phase": "process_dpo_r1_policy",
      "split": "fresh_paired",
      "mode": "learned",
      "k": 2,
      "n": 2,
      "accuracy": 0.0,
      "valid_rate": 0.0,
      "program_exact": 0.0,
      "stop_rate": 0.0,
      "false_stop_rate": 0.0,
      "mean_steps": 2.0,
      "blank_accuracy": 0.0,
      "oracle_accuracy": 0.0,
      "native_accuracy": 0.0
    },
    {
      "run": "smoke_process_dpo_20260624",
      "phase": "process_dpo_r1_policy",
      "split": "fresh_paired",
      "mode": "forced",
      "k": 2,
      "n": 2,
      "accuracy": 0.0,
      "valid_rate": 0.0,
      "program_exact": 0.0,
      "stop_rate": 0.0,
      "false_stop_rate": 0.0,
      "mean_steps": 2.0,
      "blank_accuracy": 0.0,
      "oracle_accuracy": 0.0,
      "native_accuracy": 0.0
    },
    {
      "run": "smoke_process_dpo_20260624",
      "phase": "process_dpo_r1_policy",
      "split": "hard_composition",
      "mode": "blank",
      "k": 0,
      "n": 2,
      "accuracy": 0.5,
      "valid_rate": 1.0,
      "program_exact": 0.0,
      "stop_rate": 0.0,
      "false_stop_rate": 0.0,
      "mean_steps": 0.0,
      "blank_accuracy": 0.5,
      "oracle_accuracy": 0.5,
      "native_accuracy": 0.0
    },
    {
      "run": "smoke_process_dpo_20260624",
      "phase": "process_dpo_r1_policy",
      "split": "hard_composition",
      "mode": "oracle_teacher",
      "k": 0,
      "n": 2,
      "accuracy": 0.5,
      "valid_rate": 1.0,
      "program_exact": 0.0,
      "stop_rate": 0.0,
      "false_stop_rate": 0.0,
      "mean_steps": 0.0,
      "blank_accuracy": 0.5,
      "oracle_accuracy": 0.5,
      "native_accuracy": 0.0
    },
    {
      "run": "smoke_process_dpo_20260624",
      "phase": "process_dpo_r1_policy",
      "split": "hard_composition",
      "mode": "learned",
      "k": 0,
      "n": 2,
      "accuracy": 0.5,
      "valid_rate": 1.0,
      "program_exact": 0.0,
      "stop_rate": 0.0,
      "false_stop_rate": 0.0,
      "mean_steps": 0.0,
      "blank_accuracy": 0.5,
      "oracle_accuracy": 0.5,
      "native_accuracy": 0.0
    },
    {
      "run": "smoke_process_dpo_20260624",
      "phase": "process_dpo_r1_policy",
      "split": "hard_composition",
      "mode": "forced",
      "k": 0,
      "n": 2,
      "accuracy": 0.5,
      "valid_rate": 1.0,
      "program_exact": 0.0,
      "stop_rate": 0.0,
      "false_stop_rate": 0.0,
      "mean_steps": 0.0,
      "blank_accuracy": 0.5,
      "oracle_accuracy": 0.5,
      "native_accuracy": 0.0
    },
    {
      "run": "smoke_process_dpo_20260624",
      "phase": "process_dpo_r1_policy",
      "split": "hard_composition",
      "mode": "oracle_teacher",
      "k": 2,
      "n": 2,
      "accuracy": 0.5,
      "valid_rate": 0.5,
      "program_exact": 0.0,
      "stop_rate": 0.5,
      "false_stop_rate": 0.0,
      "mean_steps": 1.0,
      "blank_accuracy": 0.5,
      "oracle_accuracy": 0.5,
      "native_accuracy": 0.0
    },
    {
      "run": "smoke_process_dpo_20260624",
      "phase": "process_dpo_r1_policy",
      "split": "hard_composition",
      "mode": "learned",
      "k": 2,
      "n": 2,
      "accuracy": 0.0,
      "valid_rate": 0.0,
      "program_exact": 0.0,
      "stop_rate": 0.0,
      "false_stop_rate": 0.0,
      "mean_steps": 2.0,
      "blank_accuracy": 0.5,
      "oracle_accuracy": 0.5,
      "native_accuracy": 0.0
    },
    {
      "run": "smoke_process_dpo_20260624",
      "phase": "process_dpo_r1_policy",
      "split": "hard_composition",
      "mode": "forced",
      "k": 2,
      "n": 2,
      "accuracy": 0.0,
      "valid_rate": 0.0,
      "program_exact": 0.0,
      "stop_rate": 0.0,
      "false_stop_rate": 0.0,
      "mean_steps": 2.0,
      "blank_accuracy": 0.5,
      "oracle_accuracy": 0.5,
      "native_accuracy": 0.0
    }
  ]
}