{
  "args": {
    "model_id": "hf-internal-testing/tiny-random-LlamaForCausalLM",
    "text_model_fallback": "Qwen/Qwen3-4B",
    "model_class": "causal",
    "load_in_4bit": false,
    "torch_dtype": "bf16",
    "device_map": "auto",
    "modulus": 17,
    "max_steps": 6,
    "train_min_len": 1,
    "train_max_len": 2,
    "answer_train_min_len": 1,
    "answer_train_max_len": 4,
    "eval_lengths": "2,4,6",
    "train_size": 24,
    "answer_train_size": 24,
    "eval_size": 16,
    "feature_batch_size": 8,
    "train_batch_size": 8,
    "eval_batch_size": 16,
    "bootstrap_steps": 4,
    "answer_steps": 4,
    "eval_every": 4,
    "stage_eval_every": 2,
    "lr": 0.001,
    "weight_decay": 0.01,
    "grad_clip": 1.0,
    "head_width": 64,
    "trace_loss_weight": 1.0,
    "executor_loss_weight": 1.0,
    "variants": "direct,compiler_trace,compiler_answer_only,compiler_trace_then_answer",
    "max_length": 512,
    "seed": 7,
    "output_dir": "experiments/qwen_trace_bootstrap_retention/runs/smoke_tiny",
    "checkpoint_dir": "",
    "verbose": false
  },
  "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,
    "transformers_loader": "AutoModelForCausalLM",
    "peft_installed": true,
    "gpu_name": "NVIDIA RTX 6000 Ada Generation",
    "gpu_vram_gb": 47.383
  },
  "hidden_dim": 16,
  "dataset": {
    "bootstrap_train_size": 24,
    "answer_train_size": 24,
    "eval_size": 16,
    "bootstrap_train_lengths": [
      1,
      2
    ],
    "answer_train_lengths": [
      1,
      4
    ],
    "eval_lengths": [
      2,
      4,
      6
    ]
  },
  "variants": {
    "direct": {
      "variant": "direct",
      "train_seconds": 0.14010047912597656,
      "checkpoints": [
        "large_artifacts/qwen_trace_bootstrap_retention/checkpoints/smoke_tiny/direct/checkpoint_final.pt"
      ],
      "train_log": [
        {
          "stage": "answer_only",
          "local_step": 1,
          "step": 1,
          "loss": 2.7840793132781982,
          "trace_loss_active": false,
          "direct_loss": 2.7840793132781982,
          "len2_direct_accuracy": 0.0625,
          "len2_direct_target_mass": 0.060011204332113266,
          "len2_executor_accuracy": NaN,
          "len2_executor_target_mass": NaN,
          "len2_init_accuracy": NaN,
          "len2_op_accuracy": NaN,
          "len2_arg_accuracy": NaN,
          "len2_program_exact": NaN,
          "len4_direct_accuracy": 0.0625,
          "len4_direct_target_mass": 0.060226909816265106,
          "len4_executor_accuracy": NaN,
          "len4_executor_target_mass": NaN,
          "len4_init_accuracy": NaN,
          "len4_op_accuracy": NaN,
          "len4_arg_accuracy": NaN,
          "len4_program_exact": NaN,
          "len6_direct_accuracy": 0.0,
          "len6_direct_target_mass": 0.05873849615454674,
          "len6_executor_accuracy": NaN,
          "len6_executor_target_mass": NaN,
          "len6_init_accuracy": NaN,
          "len6_op_accuracy": NaN,
          "len6_arg_accuracy": NaN,
          "len6_program_exact": NaN
        },
        {
          "stage": "answer_only",
          "local_step": 2,
          "step": 2,
          "loss": 2.8659682273864746,
          "trace_loss_active": false,
          "direct_loss": 2.8659682273864746,
          "len2_direct_accuracy": 0.0625,
          "len2_direct_target_mass": 0.060301460325717926,
          "len2_executor_accuracy": NaN,
          "len2_executor_target_mass": NaN,
          "len2_init_accuracy": NaN,
          "len2_op_accuracy": NaN,
          "len2_arg_accuracy": NaN,
          "len2_program_exact": NaN,
          "len4_direct_accuracy": 0.0625,
          "len4_direct_target_mass": 0.06012928485870361,
          "len4_executor_accuracy": NaN,
          "len4_executor_target_mass": NaN,
          "len4_init_accuracy": NaN,
          "len4_op_accuracy": NaN,
          "len4_arg_accuracy": NaN,
          "len4_program_exact": NaN,
          "len6_direct_accuracy": 0.0,
          "len6_direct_target_mass": 0.05858164280653,
          "len6_executor_accuracy": NaN,
          "len6_executor_target_mass": NaN,
          "len6_init_accuracy": NaN,
          "len6_op_accuracy": NaN,
          "len6_arg_accuracy": NaN,
          "len6_program_exact": NaN
        },
        {
          "stage": "answer_only",
          "local_step": 4,
          "step": 4,
          "loss": 2.8569393157958984,
          "trace_loss_active": false,
          "direct_loss": 2.8569393157958984,
          "len2_direct_accuracy": 0.0625,
          "len2_direct_target_mass": 0.060658104717731476,
          "len2_executor_accuracy": NaN,
          "len2_executor_target_mass": NaN,
          "len2_init_accuracy": NaN,
          "len2_op_accuracy": NaN,
          "len2_arg_accuracy": NaN,
          "len2_program_exact": NaN,
          "len4_direct_accuracy": 0.0625,
          "len4_direct_target_mass": 0.06020441651344299,
          "len4_executor_accuracy": NaN,
          "len4_executor_target_mass": NaN,
          "len4_init_accuracy": NaN,
          "len4_op_accuracy": NaN,
          "len4_arg_accuracy": NaN,
          "len4_program_exact": NaN,
          "len6_direct_accuracy": 0.0,
          "len6_direct_target_mass": 0.05800618231296539,
          "len6_executor_accuracy": NaN,
          "len6_executor_target_mass": NaN,
          "len6_init_accuracy": NaN,
          "len6_op_accuracy": NaN,
          "len6_arg_accuracy": NaN,
          "len6_program_exact": NaN
        },
        {
          "stage": "answer_only",
          "local_step": 6,
          "step": 6,
          "loss": 2.831756830215454,
          "trace_loss_active": false,
          "direct_loss": 2.831756830215454,
          "len2_direct_accuracy": 0.0625,
          "len2_direct_target_mass": 0.06101621314883232,
          "len2_executor_accuracy": NaN,
          "len2_executor_target_mass": NaN,
          "len2_init_accuracy": NaN,
          "len2_op_accuracy": NaN,
          "len2_arg_accuracy": NaN,
          "len2_program_exact": NaN,
          "len4_direct_accuracy": 0.0625,
          "len4_direct_target_mass": 0.059976160526275635,
          "len4_executor_accuracy": NaN,
          "len4_executor_target_mass": NaN,
          "len4_init_accuracy": NaN,
          "len4_op_accuracy": NaN,
          "len4_arg_accuracy": NaN,
          "len4_program_exact": NaN,
          "len6_direct_accuracy": 0.0,
          "len6_direct_target_mass": 0.057279251515865326,
          "len6_executor_accuracy": NaN,
          "len6_executor_target_mass": NaN,
          "len6_init_accuracy": NaN,
          "len6_op_accuracy": NaN,
          "len6_arg_accuracy": NaN,
          "len6_program_exact": NaN
        },
        {
          "stage": "answer_only",
          "local_step": 8,
          "step": 8,
          "loss": 2.823350429534912,
          "trace_loss_active": false,
          "direct_loss": 2.823350429534912,
          "len2_direct_accuracy": 0.0625,
          "len2_direct_target_mass": 0.06149920076131821,
          "len2_executor_accuracy": NaN,
          "len2_executor_target_mass": NaN,
          "len2_init_accuracy": NaN,
          "len2_op_accuracy": NaN,
          "len2_arg_accuracy": NaN,
          "len2_program_exact": NaN,
          "len4_direct_accuracy": 0.0625,
          "len4_direct_target_mass": 0.0599239356815815,
          "len4_executor_accuracy": NaN,
          "len4_executor_target_mass": NaN,
          "len4_init_accuracy": NaN,
          "len4_op_accuracy": NaN,
          "len4_arg_accuracy": NaN,
          "len4_program_exact": NaN,
          "len6_direct_accuracy": 0.0,
          "len6_direct_target_mass": 0.056633979082107544,
          "len6_executor_accuracy": NaN,
          "len6_executor_target_mass": NaN,
          "len6_init_accuracy": NaN,
          "len6_op_accuracy": NaN,
          "len6_arg_accuracy": NaN,
          "len6_program_exact": NaN
        }
      ],
      "final_metrics": {
        "len2": {
          "variant": "direct",
          "n": 16.0,
          "direct_accuracy": 0.0625,
          "direct_target_mass": 0.06149920076131821,
          "executor_accuracy": NaN,
          "executor_target_mass": NaN,
          "init_accuracy": NaN,
          "op_accuracy": NaN,
          "arg_accuracy": NaN,
          "program_exact": NaN
        },
        "len4": {
          "variant": "direct",
          "n": 16.0,
          "direct_accuracy": 0.0625,
          "direct_target_mass": 0.0599239356815815,
          "executor_accuracy": NaN,
          "executor_target_mass": NaN,
          "init_accuracy": NaN,
          "op_accuracy": NaN,
          "arg_accuracy": NaN,
          "program_exact": NaN
        },
        "len6": {
          "variant": "direct",
          "n": 16.0,
          "direct_accuracy": 0.0,
          "direct_target_mass": 0.056633979082107544,
          "executor_accuracy": NaN,
          "executor_target_mass": NaN,
          "init_accuracy": NaN,
          "op_accuracy": NaN,
          "arg_accuracy": NaN,
          "program_exact": NaN
        }
      }
    },
    "compiler_trace": {
      "variant": "compiler_trace",
      "train_seconds": 0.2516493797302246,
      "checkpoints": [
        "large_artifacts/qwen_trace_bootstrap_retention/checkpoints/smoke_tiny/compiler_trace/checkpoint_final.pt"
      ],
      "train_log": [
        {
          "stage": "trace_bootstrap",
          "local_step": 1,
          "step": 1,
          "loss": 9.611882209777832,
          "trace_loss_active": true,
          "executor_loss": 2.8636178970336914,
          "init_loss": 2.8422398567199707,
          "op_loss": 1.0785874128341675,
          "arg_loss": 2.827436923980713,
          "len2_direct_accuracy": NaN,
          "len2_direct_target_mass": NaN,
          "len2_executor_accuracy": 0.0,
          "len2_executor_target_mass": 0.06267498433589935,
          "len2_init_accuracy": 0.0,
          "len2_op_accuracy": 0.40625,
          "len2_arg_accuracy": 0.0,
          "len2_program_exact": 0.0,
          "len4_direct_accuracy": NaN,
          "len4_direct_target_mass": NaN,
          "len4_executor_accuracy": 0.125,
          "len4_executor_target_mass": 0.05694850906729698,
          "len4_init_accuracy": 0.0,
          "len4_op_accuracy": 0.296875,
          "len4_arg_accuracy": 0.0,
          "len4_program_exact": 0.0,
          "len6_direct_accuracy": NaN,
          "len6_direct_target_mass": NaN,
          "len6_executor_accuracy": 0.0625,
          "len6_executor_target_mass": 0.058530308306217194,
          "len6_init_accuracy": 0.0,
          "len6_op_accuracy": 0.3854166666666667,
          "len6_arg_accuracy": 0.0,
          "len6_program_exact": 0.0
        },
        {
          "stage": "trace_bootstrap",
          "local_step": 2,
          "step": 2,
          "loss": 9.709327697753906,
          "trace_loss_active": true,
          "executor_loss": 2.8636436462402344,
          "init_loss": 2.8307385444641113,
          "op_loss": 1.15253746509552,
          "arg_loss": 2.8624086380004883,
          "len2_direct_accuracy": NaN,
          "len2_direct_target_mass": NaN,
          "len2_executor_accuracy": 0.0,
          "len2_executor_target_mass": 0.06247786432504654,
          "len2_init_accuracy": 0.0,
          "len2_op_accuracy": 0.40625,
          "len2_arg_accuracy": 0.0,
          "len2_program_exact": 0.0,
          "len4_direct_accuracy": NaN,
          "len4_direct_target_mass": NaN,
          "len4_executor_accuracy": 0.0625,
          "len4_executor_target_mass": 0.0570644810795784,
          "len4_init_accuracy": 0.0,
          "len4_op_accuracy": 0.296875,
          "len4_arg_accuracy": 0.046875,
          "len4_program_exact": 0.0,
          "len6_direct_accuracy": NaN,
          "len6_direct_target_mass": NaN,
          "len6_executor_accuracy": 0.0625,
          "len6_executor_target_mass": 0.05854993313550949,
          "len6_init_accuracy": 0.0,
          "len6_op_accuracy": 0.3854166666666667,
          "len6_arg_accuracy": 0.07291666666666667,
          "len6_program_exact": 0.0
        },
        {
          "stage": "trace_bootstrap",
          "local_step": 4,
          "step": 4,
          "loss": 9.529363632202148,
          "trace_loss_active": true,
          "executor_loss": 2.857550621032715,
          "init_loss": 2.807183027267456,
          "op_loss": 1.0453342199325562,
          "arg_loss": 2.819295644760132,
          "len2_direct_accuracy": NaN,
          "len2_direct_target_mass": NaN,
          "len2_executor_accuracy": 0.0,
          "len2_executor_target_mass": 0.062399622052907944,
          "len2_init_accuracy": 0.125,
          "len2_op_accuracy": 0.40625,
          "len2_arg_accuracy": 0.0,
          "len2_program_exact": 0.0,
          "len4_direct_accuracy": NaN,
          "len4_direct_target_mass": NaN,
          "len4_executor_accuracy": 0.125,
          "len4_executor_target_mass": 0.05711211636662483,
          "len4_init_accuracy": 0.125,
          "len4_op_accuracy": 0.296875,
          "len4_arg_accuracy": 0.046875,
          "len4_program_exact": 0.0,
          "len6_direct_accuracy": NaN,
          "len6_direct_target_mass": NaN,
          "len6_executor_accuracy": 0.0625,
          "len6_executor_target_mass": 0.05856586992740631,
          "len6_init_accuracy": 0.125,
          "len6_op_accuracy": 0.3854166666666667,
          "len6_arg_accuracy": 0.09375,
          "len6_program_exact": 0.0
        },
        {
          "stage": "trace_continuation",
          "local_step": 1,
          "step": 5,
          "loss": 9.612869262695312,
          "trace_loss_active": true,
          "executor_loss": 2.7937378883361816,
          "init_loss": 2.8239588737487793,
          "op_loss": 1.1724225282669067,
          "arg_loss": 2.8227498531341553,
          "len2_direct_accuracy": NaN,
          "len2_direct_target_mass": NaN,
          "len2_executor_accuracy": 0.0,
          "len2_executor_target_mass": 0.06238032877445221,
          "len2_init_accuracy": 0.125,
          "len2_op_accuracy": 0.40625,
          "len2_arg_accuracy": 0.0,
          "len2_program_exact": 0.0,
          "len4_direct_accuracy": NaN,
          "len4_direct_target_mass": NaN,
          "len4_executor_accuracy": 0.125,
          "len4_executor_target_mass": 0.057130929082632065,
          "len4_init_accuracy": 0.125,
          "len4_op_accuracy": 0.296875,
          "len4_arg_accuracy": 0.046875,
          "len4_program_exact": 0.0,
          "len6_direct_accuracy": NaN,
          "len6_direct_target_mass": NaN,
          "len6_executor_accuracy": 0.0625,
          "len6_executor_target_mass": 0.05857175588607788,
          "len6_init_accuracy": 0.125,
          "len6_op_accuracy": 0.3854166666666667,
          "len6_arg_accuracy": 0.09375,
          "len6_program_exact": 0.0
        },
        {
          "stage": "trace_continuation",
          "local_step": 2,
          "step": 6,
          "loss": 9.56521987915039,
          "trace_loss_active": true,
          "executor_loss": 2.867371082305908,
          "init_loss": 2.8378758430480957,
          "op_loss": 1.1056010723114014,
          "arg_loss": 2.7543718814849854,
          "len2_direct_accuracy": NaN,
          "len2_direct_target_mass": NaN,
          "len2_executor_accuracy": 0.0,
          "len2_executor_target_mass": 0.06234161928296089,
          "len2_init_accuracy": 0.125,
          "len2_op_accuracy": 0.40625,
          "len2_arg_accuracy": 0.0,
          "len2_program_exact": 0.0,
          "len4_direct_accuracy": NaN,
          "len4_direct_target_mass": NaN,
          "len4_executor_accuracy": 0.125,
          "len4_executor_target_mass": 0.0571630634367466,
          "len4_init_accuracy": 0.125,
          "len4_op_accuracy": 0.296875,
          "len4_arg_accuracy": 0.046875,
          "len4_program_exact": 0.0,
          "len6_direct_accuracy": NaN,
          "len6_direct_target_mass": NaN,
          "len6_executor_accuracy": 0.0625,
          "len6_executor_target_mass": 0.05857925862073898,
          "len6_init_accuracy": 0.125,
          "len6_op_accuracy": 0.3854166666666667,
          "len6_arg_accuracy": 0.09375,
          "len6_program_exact": 0.0
        },
        {
          "stage": "trace_continuation",
          "local_step": 4,
          "step": 8,
          "loss": 9.461353302001953,
          "trace_loss_active": true,
          "executor_loss": 2.796470880508423,
          "init_loss": 2.791445732116699,
          "op_loss": 1.0829846858978271,
          "arg_loss": 2.790452241897583,
          "len2_direct_accuracy": NaN,
          "len2_direct_target_mass": NaN,
          "len2_executor_accuracy": 0.0,
          "len2_executor_target_mass": 0.062332119792699814,
          "len2_init_accuracy": 0.125,
          "len2_op_accuracy": 0.40625,
          "len2_arg_accuracy": 0.0,
          "len2_program_exact": 0.0,
          "len4_direct_accuracy": NaN,
          "len4_direct_target_mass": NaN,
          "len4_executor_accuracy": 0.1875,
          "len4_executor_target_mass": 0.057190753519535065,
          "len4_init_accuracy": 0.125,
          "len4_op_accuracy": 0.296875,
          "len4_arg_accuracy": 0.0625,
          "len4_program_exact": 0.0,
          "len6_direct_accuracy": NaN,
          "len6_direct_target_mass": NaN,
          "len6_executor_accuracy": 0.0625,
          "len6_executor_target_mass": 0.058585893362760544,
          "len6_init_accuracy": 0.125,
          "len6_op_accuracy": 0.3854166666666667,
          "len6_arg_accuracy": 0.10416666666666667,
          "len6_program_exact": 0.0
        }
      ],
      "final_metrics": {
        "len2": {
          "variant": "compiler_trace",
          "n": 16.0,
          "direct_accuracy": NaN,
          "direct_target_mass": NaN,
          "executor_accuracy": 0.0,
          "executor_target_mass": 0.062332119792699814,
          "init_accuracy": 0.125,
          "op_accuracy": 0.40625,
          "arg_accuracy": 0.0,
          "program_exact": 0.0
        },
        "len4": {
          "variant": "compiler_trace",
          "n": 16.0,
          "direct_accuracy": NaN,
          "direct_target_mass": NaN,
          "executor_accuracy": 0.1875,
          "executor_target_mass": 0.057190753519535065,
          "init_accuracy": 0.125,
          "op_accuracy": 0.296875,
          "arg_accuracy": 0.0625,
          "program_exact": 0.0
        },
        "len6": {
          "variant": "compiler_trace",
          "n": 16.0,
          "direct_accuracy": NaN,
          "direct_target_mass": NaN,
          "executor_accuracy": 0.0625,
          "executor_target_mass": 0.058585893362760544,
          "init_accuracy": 0.125,
          "op_accuracy": 0.3854166666666667,
          "arg_accuracy": 0.10416666666666667,
          "program_exact": 0.0
        }
      }
    },
    "compiler_answer_only": {
      "variant": "compiler_answer_only",
      "train_seconds": 0.13178706169128418,
      "checkpoints": [
        "large_artifacts/qwen_trace_bootstrap_retention/checkpoints/smoke_tiny/compiler_answer_only/checkpoint_final.pt"
      ],
      "train_log": [
        {
          "stage": "answer_only",
          "local_step": 1,
          "step": 1,
          "loss": 2.8494725227355957,
          "trace_loss_active": false,
          "executor_loss": 2.8494725227355957,
          "len2_direct_accuracy": NaN,
          "len2_direct_target_mass": NaN,
          "len2_executor_accuracy": 0.0625,
          "len2_executor_target_mass": 0.060736142098903656,
          "len2_init_accuracy": 0.0,
          "len2_op_accuracy": 0.25,
          "len2_arg_accuracy": 0.0,
          "len2_program_exact": 0.0,
          "len4_direct_accuracy": NaN,
          "len4_direct_target_mass": NaN,
          "len4_executor_accuracy": 0.0,
          "len4_executor_target_mass": 0.05797578766942024,
          "len4_init_accuracy": 0.0625,
          "len4_op_accuracy": 0.3125,
          "len4_arg_accuracy": 0.015625,
          "len4_program_exact": 0.0,
          "len6_direct_accuracy": NaN,
          "len6_direct_target_mass": NaN,
          "len6_executor_accuracy": 0.0,
          "len6_executor_target_mass": 0.05896584317088127,
          "len6_init_accuracy": 0.0,
          "len6_op_accuracy": 0.23958333333333334,
          "len6_arg_accuracy": 0.0,
          "len6_program_exact": 0.0
        },
        {
          "stage": "answer_only",
          "local_step": 2,
          "step": 2,
          "loss": 2.81935453414917,
          "trace_loss_active": false,
          "executor_loss": 2.81935453414917,
          "len2_direct_accuracy": NaN,
          "len2_direct_target_mass": NaN,
          "len2_executor_accuracy": 0.0625,
          "len2_executor_target_mass": 0.060719557106494904,
          "len2_init_accuracy": 0.0,
          "len2_op_accuracy": 0.25,
          "len2_arg_accuracy": 0.0,
          "len2_program_exact": 0.0,
          "len4_direct_accuracy": NaN,
          "len4_direct_target_mass": NaN,
          "len4_executor_accuracy": 0.0,
          "len4_executor_target_mass": 0.057977329939603806,
          "len4_init_accuracy": 0.0625,
          "len4_op_accuracy": 0.3125,
          "len4_arg_accuracy": 0.0,
          "len4_program_exact": 0.0,
          "len6_direct_accuracy": NaN,
          "len6_direct_target_mass": NaN,
          "len6_executor_accuracy": 0.0,
          "len6_executor_target_mass": 0.0589715950191021,
          "len6_init_accuracy": 0.0,
          "len6_op_accuracy": 0.23958333333333334,
          "len6_arg_accuracy": 0.0,
          "len6_program_exact": 0.0
        },
        {
          "stage": "answer_only",
          "local_step": 4,
          "step": 4,
          "loss": 2.8451733589172363,
          "trace_loss_active": false,
          "executor_loss": 2.8451733589172363,
          "len2_direct_accuracy": NaN,
          "len2_direct_target_mass": NaN,
          "len2_executor_accuracy": 0.0625,
          "len2_executor_target_mass": 0.06058919429779053,
          "len2_init_accuracy": 0.0,
          "len2_op_accuracy": 0.25,
          "len2_arg_accuracy": 0.0,
          "len2_program_exact": 0.0,
          "len4_direct_accuracy": NaN,
          "len4_direct_target_mass": NaN,
          "len4_executor_accuracy": 0.0,
          "len4_executor_target_mass": 0.05803314596414566,
          "len4_init_accuracy": 0.0625,
          "len4_op_accuracy": 0.3125,
          "len4_arg_accuracy": 0.0,
          "len4_program_exact": 0.0,
          "len6_direct_accuracy": NaN,
          "len6_direct_target_mass": NaN,
          "len6_executor_accuracy": 0.0,
          "len6_executor_target_mass": 0.05897022783756256,
          "len6_init_accuracy": 0.0,
          "len6_op_accuracy": 0.23958333333333334,
          "len6_arg_accuracy": 0.0,
          "len6_program_exact": 0.0
        },
        {
          "stage": "answer_only",
          "local_step": 6,
          "step": 6,
          "loss": 2.8455233573913574,
          "trace_loss_active": false,
          "executor_loss": 2.8455233573913574,
          "len2_direct_accuracy": NaN,
          "len2_direct_target_mass": NaN,
          "len2_executor_accuracy": 0.0625,
          "len2_executor_target_mass": 0.06051570922136307,
          "len2_init_accuracy": 0.0,
          "len2_op_accuracy": 0.25,
          "len2_arg_accuracy": 0.0,
          "len2_program_exact": 0.0,
          "len4_direct_accuracy": NaN,
          "len4_direct_target_mass": NaN,
          "len4_executor_accuracy": 0.0,
          "len4_executor_target_mass": 0.05806346982717514,
          "len4_init_accuracy": 0.0625,
          "len4_op_accuracy": 0.3125,
          "len4_arg_accuracy": 0.0,
          "len4_program_exact": 0.0,
          "len6_direct_accuracy": NaN,
          "len6_direct_target_mass": NaN,
          "len6_executor_accuracy": 0.0625,
          "len6_executor_target_mass": 0.05897083133459091,
          "len6_init_accuracy": 0.0,
          "len6_op_accuracy": 0.23958333333333334,
          "len6_arg_accuracy": 0.0,
          "len6_program_exact": 0.0
        },
        {
          "stage": "answer_only",
          "local_step": 8,
          "step": 8,
          "loss": 2.8139944076538086,
          "trace_loss_active": false,
          "executor_loss": 2.8139944076538086,
          "len2_direct_accuracy": NaN,
          "len2_direct_target_mass": NaN,
          "len2_executor_accuracy": 0.0,
          "len2_executor_target_mass": 0.06043631583452225,
          "len2_init_accuracy": 0.0,
          "len2_op_accuracy": 0.25,
          "len2_arg_accuracy": 0.0,
          "len2_program_exact": 0.0,
          "len4_direct_accuracy": NaN,
          "len4_direct_target_mass": NaN,
          "len4_executor_accuracy": 0.0,
          "len4_executor_target_mass": 0.058099739253520966,
          "len4_init_accuracy": 0.0625,
          "len4_op_accuracy": 0.3125,
          "len4_arg_accuracy": 0.0,
          "len4_program_exact": 0.0,
          "len6_direct_accuracy": NaN,
          "len6_direct_target_mass": NaN,
          "len6_executor_accuracy": 0.125,
          "len6_executor_target_mass": 0.0589713416993618,
          "len6_init_accuracy": 0.0,
          "len6_op_accuracy": 0.23958333333333334,
          "len6_arg_accuracy": 0.0,
          "len6_program_exact": 0.0
        }
      ],
      "final_metrics": {
        "len2": {
          "variant": "compiler_answer_only",
          "n": 16.0,
          "direct_accuracy": NaN,
          "direct_target_mass": NaN,
          "executor_accuracy": 0.0,
          "executor_target_mass": 0.06043631583452225,
          "init_accuracy": 0.0,
          "op_accuracy": 0.25,
          "arg_accuracy": 0.0,
          "program_exact": 0.0
        },
        "len4": {
          "variant": "compiler_answer_only",
          "n": 16.0,
          "direct_accuracy": NaN,
          "direct_target_mass": NaN,
          "executor_accuracy": 0.0,
          "executor_target_mass": 0.058099739253520966,
          "init_accuracy": 0.0625,
          "op_accuracy": 0.3125,
          "arg_accuracy": 0.0,
          "program_exact": 0.0
        },
        "len6": {
          "variant": "compiler_answer_only",
          "n": 16.0,
          "direct_accuracy": NaN,
          "direct_target_mass": NaN,
          "executor_accuracy": 0.125,
          "executor_target_mass": 0.0589713416993618,
          "init_accuracy": 0.0,
          "op_accuracy": 0.23958333333333334,
          "arg_accuracy": 0.0,
          "program_exact": 0.0
        }
      }
    },
    "compiler_trace_then_answer": {
      "variant": "compiler_trace_then_answer",
      "train_seconds": 0.14449429512023926,
      "checkpoints": [
        "large_artifacts/qwen_trace_bootstrap_retention/checkpoints/smoke_tiny/compiler_trace_then_answer/checkpoint_final.pt"
      ],
      "train_log": [
        {
          "stage": "trace_bootstrap",
          "local_step": 1,
          "step": 1,
          "loss": 9.516977310180664,
          "trace_loss_active": true,
          "executor_loss": 2.855104923248291,
          "init_loss": 2.718648910522461,
          "op_loss": 1.0843545198440552,
          "arg_loss": 2.8588693141937256,
          "len2_direct_accuracy": NaN,
          "len2_direct_target_mass": NaN,
          "len2_executor_accuracy": 0.0,
          "len2_executor_target_mass": 0.062123753130435944,
          "len2_init_accuracy": 0.125,
          "len2_op_accuracy": 0.25,
          "len2_arg_accuracy": 0.0625,
          "len2_program_exact": 0.0,
          "len4_direct_accuracy": NaN,
          "len4_direct_target_mass": NaN,
          "len4_executor_accuracy": 0.0,
          "len4_executor_target_mass": 0.05719291418790817,
          "len4_init_accuracy": 0.125,
          "len4_op_accuracy": 0.3125,
          "len4_arg_accuracy": 0.078125,
          "len4_program_exact": 0.0,
          "len6_direct_accuracy": NaN,
          "len6_direct_target_mass": NaN,
          "len6_executor_accuracy": 0.125,
          "len6_executor_target_mass": 0.05879005789756775,
          "len6_init_accuracy": 0.1875,
          "len6_op_accuracy": 0.23958333333333334,
          "len6_arg_accuracy": 0.041666666666666664,
          "len6_program_exact": 0.0
        },
        {
          "stage": "trace_bootstrap",
          "local_step": 2,
          "step": 2,
          "loss": 9.486942291259766,
          "trace_loss_active": true,
          "executor_loss": 2.8562774658203125,
          "init_loss": 2.7238242626190186,
          "op_loss": 1.0730489492416382,
          "arg_loss": 2.833791732788086,
          "len2_direct_accuracy": NaN,
          "len2_direct_target_mass": NaN,
          "len2_executor_accuracy": 0.0,
          "len2_executor_target_mass": 0.06217290461063385,
          "len2_init_accuracy": 0.125,
          "len2_op_accuracy": 0.25,
          "len2_arg_accuracy": 0.09375,
          "len2_program_exact": 0.0,
          "len4_direct_accuracy": NaN,
          "len4_direct_target_mass": NaN,
          "len4_executor_accuracy": 0.0,
          "len4_executor_target_mass": 0.05716496706008911,
          "len4_init_accuracy": 0.125,
          "len4_op_accuracy": 0.3125,
          "len4_arg_accuracy": 0.109375,
          "len4_program_exact": 0.0,
          "len6_direct_accuracy": NaN,
          "len6_direct_target_mass": NaN,
          "len6_executor_accuracy": 0.125,
          "len6_executor_target_mass": 0.05879618227481842,
          "len6_init_accuracy": 0.1875,
          "len6_op_accuracy": 0.23958333333333334,
          "len6_arg_accuracy": 0.0625,
          "len6_program_exact": 0.0
        },
        {
          "stage": "trace_bootstrap",
          "local_step": 4,
          "step": 4,
          "loss": 9.240339279174805,
          "trace_loss_active": true,
          "executor_loss": 2.857226848602295,
          "init_loss": 2.5136497020721436,
          "op_loss": 1.057813286781311,
          "arg_loss": 2.811650037765503,
          "len2_direct_accuracy": NaN,
          "len2_direct_target_mass": NaN,
          "len2_executor_accuracy": 0.0,
          "len2_executor_target_mass": 0.062321022152900696,
          "len2_init_accuracy": 0.125,
          "len2_op_accuracy": 0.25,
          "len2_arg_accuracy": 0.09375,
          "len2_program_exact": 0.0,
          "len4_direct_accuracy": NaN,
          "len4_direct_target_mass": NaN,
          "len4_executor_accuracy": 0.0625,
          "len4_executor_target_mass": 0.05708148330450058,
          "len4_init_accuracy": 0.125,
          "len4_op_accuracy": 0.3125,
          "len4_arg_accuracy": 0.140625,
          "len4_program_exact": 0.0,
          "len6_direct_accuracy": NaN,
          "len6_direct_target_mass": NaN,
          "len6_executor_accuracy": 0.125,
          "len6_executor_target_mass": 0.058807067573070526,
          "len6_init_accuracy": 0.1875,
          "len6_op_accuracy": 0.23958333333333334,
          "len6_arg_accuracy": 0.10416666666666667,
          "len6_program_exact": 0.0
        },
        {
          "stage": "answer_retention",
          "local_step": 1,
          "step": 5,
          "loss": 2.769686698913574,
          "trace_loss_active": false,
          "executor_loss": 2.769686698913574,
          "len2_direct_accuracy": NaN,
          "len2_direct_target_mass": NaN,
          "len2_executor_accuracy": 0.0,
          "len2_executor_target_mass": 0.062424324452877045,
          "len2_init_accuracy": 0.125,
          "len2_op_accuracy": 0.25,
          "len2_arg_accuracy": 0.125,
          "len2_program_exact": 0.0,
          "len4_direct_accuracy": NaN,
          "len4_direct_target_mass": NaN,
          "len4_executor_accuracy": 0.0,
          "len4_executor_target_mass": 0.057016514241695404,
          "len4_init_accuracy": 0.125,
          "len4_op_accuracy": 0.3125,
          "len4_arg_accuracy": 0.21875,
          "len4_program_exact": 0.0,
          "len6_direct_accuracy": NaN,
          "len6_direct_target_mass": NaN,
          "len6_executor_accuracy": 0.0625,
          "len6_executor_target_mass": 0.05881367623806,
          "len6_init_accuracy": 0.1875,
          "len6_op_accuracy": 0.23958333333333334,
          "len6_arg_accuracy": 0.14583333333333334,
          "len6_program_exact": 0.0
        },
        {
          "stage": "answer_retention",
          "local_step": 2,
          "step": 6,
          "loss": 2.8157098293304443,
          "trace_loss_active": false,
          "executor_loss": 2.8157098293304443,
          "len2_direct_accuracy": NaN,
          "len2_direct_target_mass": NaN,
          "len2_executor_accuracy": 0.0,
          "len2_executor_target_mass": 0.06253434717655182,
          "len2_init_accuracy": 0.1875,
          "len2_op_accuracy": 0.25,
          "len2_arg_accuracy": 0.1875,
          "len2_program_exact": 0.0,
          "len4_direct_accuracy": NaN,
          "len4_direct_target_mass": NaN,
          "len4_executor_accuracy": 0.0625,
          "len4_executor_target_mass": 0.056950267404317856,
          "len4_init_accuracy": 0.125,
          "len4_op_accuracy": 0.40625,
          "len4_arg_accuracy": 0.21875,
          "len4_program_exact": 0.0,
          "len6_direct_accuracy": NaN,
          "len6_direct_target_mass": NaN,
          "len6_executor_accuracy": 0.0,
          "len6_executor_target_mass": 0.05881597101688385,
          "len6_init_accuracy": 0.25,
          "len6_op_accuracy": 0.25,
          "len6_arg_accuracy": 0.15625,
          "len6_program_exact": 0.0
        },
        {
          "stage": "answer_retention",
          "local_step": 4,
          "step": 8,
          "loss": 2.859210252761841,
          "trace_loss_active": false,
          "executor_loss": 2.859210252761841,
          "len2_direct_accuracy": NaN,
          "len2_direct_target_mass": NaN,
          "len2_executor_accuracy": 0.125,
          "len2_executor_target_mass": 0.06274983286857605,
          "len2_init_accuracy": 0.1875,
          "len2_op_accuracy": 0.40625,
          "len2_arg_accuracy": 0.1875,
          "len2_program_exact": 0.0,
          "len4_direct_accuracy": NaN,
          "len4_direct_target_mass": NaN,
          "len4_executor_accuracy": 0.0,
          "len4_executor_target_mass": 0.05681289732456207,
          "len4_init_accuracy": 0.125,
          "len4_op_accuracy": 0.296875,
          "len4_arg_accuracy": 0.21875,
          "len4_program_exact": 0.0,
          "len6_direct_accuracy": NaN,
          "len6_direct_target_mass": NaN,
          "len6_executor_accuracy": 0.0625,
          "len6_executor_target_mass": 0.05882197618484497,
          "len6_init_accuracy": 0.25,
          "len6_op_accuracy": 0.3854166666666667,
          "len6_arg_accuracy": 0.15625,
          "len6_program_exact": 0.0
        }
      ],
      "final_metrics": {
        "len2": {
          "variant": "compiler_trace_then_answer",
          "n": 16.0,
          "direct_accuracy": NaN,
          "direct_target_mass": NaN,
          "executor_accuracy": 0.125,
          "executor_target_mass": 0.06274983286857605,
          "init_accuracy": 0.1875,
          "op_accuracy": 0.40625,
          "arg_accuracy": 0.1875,
          "program_exact": 0.0
        },
        "len4": {
          "variant": "compiler_trace_then_answer",
          "n": 16.0,
          "direct_accuracy": NaN,
          "direct_target_mass": NaN,
          "executor_accuracy": 0.0,
          "executor_target_mass": 0.05681289732456207,
          "init_accuracy": 0.125,
          "op_accuracy": 0.296875,
          "arg_accuracy": 0.21875,
          "program_exact": 0.0
        },
        "len6": {
          "variant": "compiler_trace_then_answer",
          "n": 16.0,
          "direct_accuracy": NaN,
          "direct_target_mass": NaN,
          "executor_accuracy": 0.0625,
          "executor_target_mass": 0.05882197618484497,
          "init_accuracy": 0.25,
          "op_accuracy": 0.3854166666666667,
          "arg_accuracy": 0.15625,
          "program_exact": 0.0
        }
      }
    }
  }
}