{
  "schema_version": 2,
  "adapter_key": "grpo-difficulty-curriculum",
  "manifest_ref": "reproduction:grpo-difficulty-curriculum",
  "method": "grpo-difficulty-curriculum",
  "measured_locally": true,
  "diagnostic_only": true,
  "promotion_eligible": false,
  "seeds": [
    42
  ],
  "seed_results": [
    {
      "seed": 42,
      "diagnostic_only": true,
      "device": "cuda",
      "variants": {
        "fixed-medium": {
          "method": "grpo-difficulty-curriculum",
          "seed": 42,
          "device": "cuda",
          "diagnostic_only": true,
          "promotion_eligible": false,
          "evaluation_tier": "l1_mechanism",
          "dataset": {
            "verified": true,
            "test_read": false,
            "source_id": "openai/grade-school-math",
            "source_revision": "3101c7d5072418e28b9008a6636bde82a006892c",
            "source_sha256": "17f347dc51477c50d4efb83959dbb7c56297aba886e5544ee2aaed3024813465",
            "partition_sha256": {
              "train": "6b8672439c8cf6f6b6e65bf5f0e515fbf4c497c9066eefb62a6a6a61ced158e3",
              "validation": "2cf1119ad7b2e31e825274487ef32c073e03a8ed44f08ecd3e0366c1f40ebdb7"
            },
            "partition": "normalized-question sha256 unsigned-prefix / 2^64 < 0.1 validation"
          },
          "checkpoint": {
            "public_id": "Qwen/Qwen2.5-0.5B-Instruct",
            "revision": "7ae557604adf67be50417f59c2c2f167def9a775",
            "verified_mirror": {
              "public_id": "Qwen/Qwen2.5-0.5B-Instruct",
              "revision": "7ae557604adf67be50417f59c2c2f167def9a775",
              "file_sha256": {
                "config.json": "18e18afcaccafade98daf13a54092927904649e1dd4eba8299ab717d5d94ff45",
                "generation_config.json": "e558847a8b4402616f1273797b015104dc266fe4b520056fca88823ba8f8ebe6",
                "merges.txt": "599bab54075088774b1733fde865d5bd747cbcc7a547c5bc12610e874e26f5e3",
                "model.safetensors": "fdf756fa7fcbe7404d5c60e26bff1a0c8b8aa1f72ced49e7dd0210fe288fb7fe",
                "tokenizer.json": "c0382117ea329cdf097041132f6d735924b697924d6f6fc3945713e96ce87539",
                "tokenizer_config.json": "5b5d4f65d0acd3b2d56a35b56d374a36cbc1c8fa5cf3b3febbbfabf22f359583",
                "vocab.json": "ca10d7e9fb3ed18575dd1e277a2579c16d108e32f27439684afa0e10b1440910"
              }
            }
          },
          "baseline": {
            "accuracy": 0.5,
            "examples": 4,
            "generated_tokens": 1419,
            "truncated_responses": 1
          },
          "method_metrics": {
            "accuracy": 0.75,
            "examples": 4,
            "generated_tokens": 1263,
            "truncated_responses": 0
          },
          "training": [
            {
              "step": 1,
              "band": [
                0.375,
                0.625
              ],
              "groups": [
                {
                  "training_row": 0,
                  "eligible": false,
                  "current_success": 0.75
                },
                {
                  "training_row": 1,
                  "eligible": true,
                  "current_success": 0.5,
                  "correctness_rewards": [
                    0,
                    1,
                    0,
                    1,
                    1,
                    0,
                    1,
                    0
                  ],
                  "loss": -1.9371509552001953e-07,
                  "gradient_l2_norm": 0.15984754264354706,
                  "optimizer_update_executed": true,
                  "zero_advantage_group": false,
                  "rewards": [
                    0.2,
                    1.2,
                    0.2,
                    1.2,
                    1.2,
                    0.2,
                    1.2,
                    0.2
                  ],
                  "advantages": [
                    -0.9999978542327881,
                    0.9999982118606567,
                    -0.9999978542327881,
                    0.9999982118606567,
                    0.9999982118606567,
                    -0.9999978542327881,
                    0.9999982118606567,
                    -0.9999978542327881
                  ]
                },
                {
                  "training_row": 2,
                  "eligible": false,
                  "current_success": 0.25
                },
                {
                  "training_row": 3,
                  "eligible": false,
                  "current_success": 0.25
                },
                {
                  "training_row": 4,
                  "eligible": false,
                  "current_success": 0
                },
                {
                  "training_row": 5,
                  "eligible": false,
                  "current_success": 0.75
                },
                {
                  "training_row": 6,
                  "eligible": false,
                  "current_success": 0
                },
                {
                  "training_row": 7,
                  "eligible": false,
                  "current_success": 0
                }
              ],
              "validation_diagnostic": {
                "accuracy": 0.25,
                "examples": 4,
                "generated_tokens": 1434,
                "truncated_responses": 0
              },
              "transition": "not-staged",
              "effective_updates": 1
            },
            {
              "step": 2,
              "band": [
                0.375,
                0.625
              ],
              "groups": [
                {
                  "training_row": 0,
                  "eligible": false,
                  "current_success": 0.875
                },
                {
                  "training_row": 1,
                  "eligible": true,
                  "current_success": 0.5,
                  "correctness_rewards": [
                    1,
                    0,
                    0,
                    1,
                    1,
                    0,
                    0,
                    1
                  ],
                  "loss": -1.9371509552001953e-07,
                  "gradient_l2_norm": 0.1646161526441574,
                  "optimizer_update_executed": true,
                  "zero_advantage_group": false,
                  "rewards": [
                    1.2,
                    0.2,
                    0.2,
                    1.2,
                    1.2,
                    0.2,
                    0.2,
                    1.2
                  ],
                  "advantages": [
                    0.9999982118606567,
                    -0.9999978542327881,
                    -0.9999978542327881,
                    0.9999982118606567,
                    0.9999982118606567,
                    -0.9999978542327881,
                    -0.9999978542327881,
                    0.9999982118606567
                  ]
                },
                {
                  "training_row": 2,
                  "eligible": false,
                  "current_success": 0.25
                },
                {
                  "training_row": 3,
                  "eligible": true,
                  "current_success": 0.375,
                  "correctness_rewards": [
                    0,
                    1,
                    0,
                    0,
                    0,
                    1,
                    0,
                    1
                  ],
                  "loss": -0.00014882534742355347,
                  "gradient_l2_norm": 0.16405336558818817,
                  "optimizer_update_executed": true,
                  "zero_advantage_group": false,
                  "rewards": [
                    0.2,
                    1.2,
                    0.2,
                    0.2,
                    0.2,
                    1.2,
                    0.2,
                    1.2
                  ],
                  "advantages": [
                    -0.7745951414108276,
                    1.2909916639328003,
                    -0.7745951414108276,
                    -0.7745951414108276,
                    -0.7745951414108276,
                    1.2909916639328003,
                    -0.7745951414108276,
                    1.2909916639328003
                  ]
                },
                {
                  "training_row": 4,
                  "eligible": false,
                  "current_success": 0
                },
                {
                  "training_row": 5,
                  "eligible": false,
                  "current_success": 1
                },
                {
                  "training_row": 6,
                  "eligible": false,
                  "current_success": 0
                },
                {
                  "training_row": 7,
                  "eligible": false,
                  "current_success": 0
                }
              ],
              "validation_diagnostic": {
                "accuracy": 0.5,
                "examples": 4,
                "generated_tokens": 1663,
                "truncated_responses": 1
              },
              "transition": "not-staged",
              "effective_updates": 2
            },
            {
              "step": 3,
              "band": [
                0.375,
                0.625
              ],
              "groups": [
                {
                  "training_row": 0,
                  "eligible": false,
                  "current_success": 0.75
                },
                {
                  "training_row": 1,
                  "eligible": true,
                  "current_success": 0.5,
                  "correctness_rewards": [
                    0,
                    1,
                    0,
                    0,
                    1,
                    1,
                    0,
                    1
                  ],
                  "loss": 5.960464477539063e-08,
                  "gradient_l2_norm": 0.16914469003677368,
                  "optimizer_update_executed": true,
                  "zero_advantage_group": false,
                  "rewards": [
                    0.2,
                    1.2,
                    0.2,
                    0.2,
                    1.2,
                    1.2,
                    0.2,
                    1.2
                  ],
                  "advantages": [
                    -0.9999980926513672,
                    0.9999979734420776,
                    -0.9999980926513672,
                    -0.9999980926513672,
                    0.9999979734420776,
                    0.9999979734420776,
                    -0.9999980926513672,
                    0.9999979734420776
                  ]
                },
                {
                  "training_row": 2,
                  "eligible": false,
                  "current_success": 0.125
                },
                {
                  "training_row": 3,
                  "eligible": true,
                  "current_success": 0.625,
                  "correctness_rewards": [
                    0,
                    1,
                    1,
                    1,
                    1,
                    0,
                    0,
                    1
                  ],
                  "loss": -0.0013230517506599426,
                  "gradient_l2_norm": 0.22981058061122894,
                  "optimizer_update_executed": true,
                  "zero_advantage_group": false,
                  "rewards": [
                    0.2,
                    1.2,
                    1.2,
                    1.2,
                    1.2,
                    0.2,
                    0.2,
                    1.2
                  ],
                  "advantages": [
                    -1.2909915447235107,
                    0.7745952606201172,
                    0.7745952606201172,
                    0.7745952606201172,
                    0.7745952606201172,
                    -1.2909915447235107,
                    -1.2909915447235107,
                    0.7745952606201172
                  ]
                },
                {
                  "training_row": 4,
                  "eligible": false,
                  "current_success": 0
                },
                {
                  "training_row": 5,
                  "eligible": false,
                  "current_success": 0.75
                },
                {
                  "training_row": 6,
                  "eligible": false,
                  "current_success": 0.125
                },
                {
                  "training_row": 7,
                  "eligible": false,
                  "current_success": 0.25
                }
              ],
              "validation_diagnostic": {
                "accuracy": 0.25,
                "examples": 4,
                "generated_tokens": 1371,
                "truncated_responses": 0
              },
              "transition": "not-staged",
              "effective_updates": 2
            },
            {
              "step": 4,
              "band": [
                0.375,
                0.625
              ],
              "groups": [
                {
                  "training_row": 0,
                  "eligible": false,
                  "current_success": 1
                },
                {
                  "training_row": 1,
                  "eligible": true,
                  "current_success": 0.625,
                  "correctness_rewards": [
                    1,
                    1,
                    0,
                    1,
                    0,
                    1,
                    0,
                    1
                  ],
                  "loss": -2.1606683731079102e-07,
                  "gradient_l2_norm": 0.15342898666858673,
                  "optimizer_update_executed": true,
                  "zero_advantage_group": false,
                  "rewards": [
                    1.2,
                    1.2,
                    0.2,
                    1.2,
                    0.2,
                    1.2,
                    0.2,
                    1.2
                  ],
                  "advantages": [
                    0.7745952606201172,
                    0.7745952606201172,
                    -1.2909915447235107,
                    0.7745952606201172,
                    -1.2909915447235107,
                    0.7745952606201172,
                    -1.2909915447235107,
                    0.7745952606201172
                  ]
                },
                {
                  "training_row": 2,
                  "eligible": false,
                  "current_success": 0.125
                },
                {
                  "training_row": 3,
                  "eligible": true,
                  "current_success": 0.375,
                  "correctness_rewards": [
                    0,
                    1,
                    0,
                    0,
                    0,
                    1,
                    1,
                    0
                  ],
                  "loss": 0.0008457452058792114,
                  "gradient_l2_norm": 0.17083901166915894,
                  "optimizer_update_executed": true,
                  "zero_advantage_group": false,
                  "rewards": [
                    0.2,
                    1.2,
                    0.2,
                    0.2,
                    0.2,
                    1.2,
                    1.2,
                    0.2
                  ],
                  "advantages": [
                    -0.7745951414108276,
                    1.2909916639328003,
                    -0.7745951414108276,
                    -0.7745951414108276,
                    -0.7745951414108276,
                    1.2909916639328003,
                    1.2909916639328003,
                    -0.7745951414108276
                  ]
                },
                {
                  "training_row": 4,
                  "eligible": false,
                  "current_success": 0.125
                },
                {
                  "training_row": 5,
                  "eligible": true,
                  "current_success": 0.625,
                  "correctness_rewards": [
                    0,
                    1,
                    0,
                    0,
                    1,
                    1,
                    1,
                    1
                  ],
                  "loss": 0.0008641406893730164,
                  "gradient_l2_norm": 0.18051153421401978,
                  "optimizer_update_executed": true,
                  "zero_advantage_group": false,
                  "rewards": [
                    0.2,
                    1.2,
                    0.2,
                    0.2,
                    1.2,
                    1.2,
                    1.2,
                    1.2
                  ],
                  "advantages": [
                    -1.2909917831420898,
                    0.7745949625968933,
                    -1.2909917831420898,
                    -1.2909917831420898,
                    0.7745949625968933,
                    0.7745949625968933,
                    0.7745949625968933,
                    0.7745949625968933
                  ]
                },
                {
                  "training_row": 6,
                  "eligible": false,
                  "current_success": 0
                },
                {
                  "training_row": 7,
                  "eligible": false,
                  "current_success": 0
                }
              ],
              "validation_diagnostic": {
                "accuracy": 0.75,
                "examples": 4,
                "generated_tokens": 1263,
                "truncated_responses": 0
              },
              "transition": "not-staged",
              "effective_updates": 3
            }
          ],
          "cost": {
            "generation_calls": {
              "evaluation": 24,
              "student": 32
            },
            "generated_tokens": {
              "evaluation": 8413,
              "student": 72642
            },
            "scored_response_tokens": {
              "behavior": 72642,
              "student_rl": 18142
            },
            "frozen_encoder_forward_calls": 0,
            "frozen_encoder_input_tokens_including_prompt": 0,
            "scope": "entire invocation, including training controls, profiling and evaluation",
            "includes_rejected_and_profiling_generations": true
          },
          "protocol": {
            "test_opened": false,
            "gold_in_student_generation": false,
            "train_examples": 8,
            "validation_examples": 4,
            "max_tokens": 512
          },
          "parameters_changed": true,
          "variant": "fixed-medium",
          "difficulty_history": [
            [
              0.75,
              0.5,
              0.25,
              0.25,
              0,
              0.75,
              0,
              0
            ],
            [
              0.875,
              0.5,
              0.25,
              0.375,
              0,
              1,
              0,
              0
            ],
            [
              0.75,
              0.5,
              0.125,
              0.625,
              0,
              0.75,
              0.125,
              0.25
            ],
            [
              1,
              0.625,
              0.125,
              0.375,
              0.125,
              0.625,
              0,
              0
            ]
          ],
          "difficulty_dynamics": {
            "span": [
              0.25,
              0.125,
              0.125,
              0.375,
              0.125,
              0.375,
              0.125,
              0.25
            ],
            "total_variation": [
              0.5,
              0.125,
              0.125,
              0.625,
              0.125,
              0.625,
              0.25,
              0.5
            ],
            "extra_round_trip_variation": [
              0.25,
              0,
              0,
              0.25,
              0,
              0.25,
              0.125,
              0.25
            ]
          },
          "initial_static_easy_to_hard_order": null,
          "execution_status": "passed",
          "limitations": [
            "Small public train-only GSM8K holdout diagnostic; no official test is read.",
            "Qwen2.5-0.5B replaces 1.5B and diagnostic cadence is scaled to each bounded step.",
            "Run fixed-medium/staged/too-hard/unfiltered/static separately with matched configured budgets.",
            "All rejected groups and calibration generations are charged; no invented nonzero advantage.",
            "Staged diagnostic uses the validation split, not original 128-per-band training-heldout pools."
          ]
        },
        "staged": {
          "method": "grpo-difficulty-curriculum",
          "seed": 42,
          "device": "cuda",
          "diagnostic_only": true,
          "promotion_eligible": false,
          "evaluation_tier": "l1_mechanism",
          "dataset": {
            "verified": true,
            "test_read": false,
            "source_id": "openai/grade-school-math",
            "source_revision": "3101c7d5072418e28b9008a6636bde82a006892c",
            "source_sha256": "17f347dc51477c50d4efb83959dbb7c56297aba886e5544ee2aaed3024813465",
            "partition_sha256": {
              "train": "6b8672439c8cf6f6b6e65bf5f0e515fbf4c497c9066eefb62a6a6a61ced158e3",
              "validation": "2cf1119ad7b2e31e825274487ef32c073e03a8ed44f08ecd3e0366c1f40ebdb7"
            },
            "partition": "normalized-question sha256 unsigned-prefix / 2^64 < 0.1 validation"
          },
          "checkpoint": {
            "public_id": "Qwen/Qwen2.5-0.5B-Instruct",
            "revision": "7ae557604adf67be50417f59c2c2f167def9a775",
            "verified_mirror": {
              "public_id": "Qwen/Qwen2.5-0.5B-Instruct",
              "revision": "7ae557604adf67be50417f59c2c2f167def9a775",
              "file_sha256": {
                "config.json": "18e18afcaccafade98daf13a54092927904649e1dd4eba8299ab717d5d94ff45",
                "generation_config.json": "e558847a8b4402616f1273797b015104dc266fe4b520056fca88823ba8f8ebe6",
                "merges.txt": "599bab54075088774b1733fde865d5bd747cbcc7a547c5bc12610e874e26f5e3",
                "model.safetensors": "fdf756fa7fcbe7404d5c60e26bff1a0c8b8aa1f72ced49e7dd0210fe288fb7fe",
                "tokenizer.json": "c0382117ea329cdf097041132f6d735924b697924d6f6fc3945713e96ce87539",
                "tokenizer_config.json": "5b5d4f65d0acd3b2d56a35b56d374a36cbc1c8fa5cf3b3febbbfabf22f359583",
                "vocab.json": "ca10d7e9fb3ed18575dd1e277a2579c16d108e32f27439684afa0e10b1440910"
              }
            }
          },
          "baseline": {
            "accuracy": 0.5,
            "examples": 4,
            "generated_tokens": 1419,
            "truncated_responses": 1
          },
          "method_metrics": {
            "accuracy": 0.5,
            "examples": 4,
            "generated_tokens": 1512,
            "truncated_responses": 1
          },
          "training": [
            {
              "step": 1,
              "band": [
                0.625,
                0.875
              ],
              "groups": [
                {
                  "training_row": 0,
                  "eligible": true,
                  "current_success": 0.75,
                  "correctness_rewards": [
                    0,
                    1,
                    1,
                    1,
                    1,
                    1,
                    1,
                    0
                  ],
                  "loss": -2.5331974029541016e-07,
                  "gradient_l2_norm": 0.2029796838760376,
                  "optimizer_update_executed": true,
                  "zero_advantage_group": false,
                  "rewards": [
                    0.2,
                    1.2,
                    1.2,
                    1.2,
                    1.2,
                    1.2,
                    1.2,
                    0.2
                  ],
                  "advantages": [
                    -1.7320466041564941,
                    0.5773491859436035,
                    0.5773491859436035,
                    0.5773491859436035,
                    0.5773491859436035,
                    0.5773491859436035,
                    0.5773491859436035,
                    -1.7320466041564941
                  ]
                },
                {
                  "training_row": 1,
                  "eligible": false,
                  "current_success": 0.5
                },
                {
                  "training_row": 2,
                  "eligible": false,
                  "current_success": 0.25
                },
                {
                  "training_row": 3,
                  "eligible": false,
                  "current_success": 0.25
                },
                {
                  "training_row": 4,
                  "eligible": false,
                  "current_success": 0
                },
                {
                  "training_row": 5,
                  "eligible": true,
                  "current_success": 0.75,
                  "correctness_rewards": [
                    0,
                    1,
                    1,
                    1,
                    1,
                    1,
                    0,
                    1
                  ],
                  "loss": 0.0009970217943191528,
                  "gradient_l2_norm": 0.15946151316165924,
                  "optimizer_update_executed": true,
                  "zero_advantage_group": false,
                  "rewards": [
                    0.2,
                    1.2,
                    1.2,
                    1.2,
                    1.2,
                    1.2,
                    0.2,
                    1.2
                  ],
                  "advantages": [
                    -1.7320464849472046,
                    0.5773491859436035,
                    0.5773491859436035,
                    0.5773491859436035,
                    0.5773491859436035,
                    0.5773491859436035,
                    -1.7320464849472046,
                    0.5773491859436035
                  ]
                },
                {
                  "training_row": 6,
                  "eligible": false,
                  "current_success": 0
                },
                {
                  "training_row": 7,
                  "eligible": false,
                  "current_success": 0
                }
              ],
              "validation_diagnostic": {
                "accuracy": 0.5,
                "examples": 4,
                "generated_tokens": 1561,
                "truncated_responses": 0
              },
              "transition": "new-best",
              "effective_updates": 2
            },
            {
              "step": 2,
              "band": [
                0.625,
                0.875
              ],
              "groups": [
                {
                  "training_row": 0,
                  "eligible": true,
                  "current_success": 0.875,
                  "correctness_rewards": [
                    1,
                    1,
                    1,
                    0,
                    1,
                    1,
                    1,
                    1
                  ],
                  "loss": -3.501772880554199e-07,
                  "gradient_l2_norm": 0.20516759157180786,
                  "optimizer_update_executed": true,
                  "zero_advantage_group": false,
                  "rewards": [
                    1.2,
                    1.2,
                    1.2,
                    0.2,
                    1.2,
                    1.2,
                    1.2,
                    1.2
                  ],
                  "advantages": [
                    0.37796369194984436,
                    0.37796369194984436,
                    0.37796369194984436,
                    -2.6457431316375732,
                    0.37796369194984436,
                    0.37796369194984436,
                    0.37796369194984436,
                    0.37796369194984436
                  ]
                },
                {
                  "training_row": 1,
                  "eligible": false,
                  "current_success": 0.25
                },
                {
                  "training_row": 2,
                  "eligible": false,
                  "current_success": 0
                },
                {
                  "training_row": 3,
                  "eligible": true,
                  "current_success": 0.625,
                  "correctness_rewards": [
                    1,
                    1,
                    1,
                    0,
                    0,
                    1,
                    1,
                    0
                  ],
                  "loss": 0.0009714290499687195,
                  "gradient_l2_norm": 0.1547374576330185,
                  "optimizer_update_executed": true,
                  "zero_advantage_group": false,
                  "rewards": [
                    1.2,
                    1.2,
                    1.2,
                    0.2,
                    0.2,
                    1.2,
                    1.2,
                    0.2
                  ],
                  "advantages": [
                    0.7745952606201172,
                    0.7745952606201172,
                    0.7745952606201172,
                    -1.2909915447235107,
                    -1.2909915447235107,
                    0.7745952606201172,
                    0.7745952606201172,
                    -1.2909915447235107
                  ]
                },
                {
                  "training_row": 4,
                  "eligible": false,
                  "current_success": 0.125
                },
                {
                  "training_row": 5,
                  "eligible": true,
                  "current_success": 0.75,
                  "correctness_rewards": [
                    0,
                    1,
                    0,
                    1,
                    1,
                    1,
                    1,
                    1
                  ],
                  "loss": -0.0003462955355644226,
                  "gradient_l2_norm": 0.17075270414352417,
                  "optimizer_update_executed": true,
                  "zero_advantage_group": false,
                  "rewards": [
                    0.2,
                    1.2,
                    0.2,
                    1.2,
                    1.2,
                    1.2,
                    1.2,
                    1.2
                  ],
                  "advantages": [
                    -1.7320464849472046,
                    0.5773491859436035,
                    -1.7320464849472046,
                    0.5773491859436035,
                    0.5773491859436035,
                    0.5773491859436035,
                    0.5773491859436035,
                    0.5773491859436035
                  ]
                },
                {
                  "training_row": 6,
                  "eligible": false,
                  "current_success": 0
                },
                {
                  "training_row": 7,
                  "eligible": false,
                  "current_success": 0.125
                }
              ],
              "validation_diagnostic": {
                "accuracy": 0.5,
                "examples": 4,
                "generated_tokens": 1562,
                "truncated_responses": 1
              },
              "transition": "keep",
              "effective_updates": 3
            },
            {
              "step": 3,
              "band": [
                0.625,
                0.875
              ],
              "groups": [
                {
                  "training_row": 0,
                  "eligible": false,
                  "current_success": 1
                },
                {
                  "training_row": 1,
                  "eligible": false,
                  "current_success": 0.375
                },
                {
                  "training_row": 2,
                  "eligible": false,
                  "current_success": 0.125
                },
                {
                  "training_row": 3,
                  "eligible": true,
                  "current_success": 0.75,
                  "correctness_rewards": [
                    1,
                    0,
                    1,
                    1,
                    1,
                    1,
                    0,
                    1
                  ],
                  "loss": -2.384185791015625e-07,
                  "gradient_l2_norm": 0.16247574985027313,
                  "optimizer_update_executed": true,
                  "zero_advantage_group": false,
                  "rewards": [
                    1.2,
                    0.2,
                    1.2,
                    1.2,
                    1.2,
                    1.2,
                    0.2,
                    1.2
                  ],
                  "advantages": [
                    0.5773491859436035,
                    -1.7320466041564941,
                    0.5773491859436035,
                    0.5773491859436035,
                    0.5773491859436035,
                    0.5773491859436035,
                    -1.7320466041564941,
                    0.5773491859436035
                  ]
                },
                {
                  "training_row": 4,
                  "eligible": false,
                  "current_success": 0.125
                },
                {
                  "training_row": 5,
                  "eligible": true,
                  "current_success": 0.75,
                  "correctness_rewards": [
                    1,
                    0,
                    0,
                    1,
                    1,
                    1,
                    1,
                    1
                  ],
                  "loss": 0.0006367117166519165,
                  "gradient_l2_norm": 0.16295728087425232,
                  "optimizer_update_executed": true,
                  "zero_advantage_group": false,
                  "rewards": [
                    1.2,
                    0.2,
                    0.2,
                    1.2,
                    1.2,
                    1.2,
                    1.2,
                    1.2
                  ],
                  "advantages": [
                    0.5773491859436035,
                    -1.7320466041564941,
                    -1.7320466041564941,
                    0.5773491859436035,
                    0.5773491859436035,
                    0.5773491859436035,
                    0.5773491859436035,
                    0.5773491859436035
                  ]
                },
                {
                  "training_row": 6,
                  "eligible": false,
                  "current_success": 0
                },
                {
                  "training_row": 7,
                  "eligible": false,
                  "current_success": 0.125
                }
              ],
              "validation_diagnostic": {
                "accuracy": 0.5,
                "examples": 4,
                "generated_tokens": 1592,
                "truncated_responses": 0
              },
              "transition": "advance",
              "effective_updates": 2
            },
            {
              "step": 4,
              "band": [
                0.5,
                0.75
              ],
              "groups": [
                {
                  "training_row": 0,
                  "eligible": false,
                  "current_success": 0.875
                },
                {
                  "training_row": 1,
                  "eligible": true,
                  "current_success": 0.5,
                  "correctness_rewards": [
                    0,
                    1,
                    0,
                    0,
                    0,
                    1,
                    1,
                    1
                  ],
                  "loss": 9.685754776000977e-08,
                  "gradient_l2_norm": 0.216303288936615,
                  "optimizer_update_executed": true,
                  "zero_advantage_group": false,
                  "rewards": [
                    0.2,
                    1.2,
                    0.2,
                    0.2,
                    0.2,
                    1.2,
                    1.2,
                    1.2
                  ],
                  "advantages": [
                    -0.9999980926513672,
                    0.9999979734420776,
                    -0.9999980926513672,
                    -0.9999980926513672,
                    -0.9999980926513672,
                    0.9999979734420776,
                    0.9999979734420776,
                    0.9999979734420776
                  ]
                },
                {
                  "training_row": 2,
                  "eligible": false,
                  "current_success": 0
                },
                {
                  "training_row": 3,
                  "eligible": true,
                  "current_success": 0.625,
                  "correctness_rewards": [
                    1,
                    1,
                    1,
                    0,
                    0,
                    1,
                    1,
                    0
                  ],
                  "loss": -0.000855766236782074,
                  "gradient_l2_norm": 0.15474621951580048,
                  "optimizer_update_executed": true,
                  "zero_advantage_group": false,
                  "rewards": [
                    1.2,
                    1.2,
                    1.2,
                    0.2,
                    0.2,
                    1.2,
                    1.2,
                    0.2
                  ],
                  "advantages": [
                    0.7745952606201172,
                    0.7745952606201172,
                    0.7745952606201172,
                    -1.2909915447235107,
                    -1.2909915447235107,
                    0.7745952606201172,
                    0.7745952606201172,
                    -1.2909915447235107
                  ]
                },
                {
                  "training_row": 4,
                  "eligible": false,
                  "current_success": 0
                },
                {
                  "training_row": 5,
                  "eligible": false,
                  "current_success": 0.875
                },
                {
                  "training_row": 6,
                  "eligible": false,
                  "current_success": 0.25
                },
                {
                  "training_row": 7,
                  "eligible": false,
                  "current_success": 0
                }
              ],
              "validation_diagnostic": {
                "accuracy": 0.5,
                "examples": 4,
                "generated_tokens": 1512,
                "truncated_responses": 1
              },
              "transition": "new-best",
              "effective_updates": 2
            }
          ],
          "cost": {
            "generation_calls": {
              "evaluation": 24,
              "student": 32
            },
            "generated_tokens": {
              "evaluation": 9158,
              "student": 73273
            },
            "scored_response_tokens": {
              "behavior": 73273,
              "student_rl": 18844
            },
            "frozen_encoder_forward_calls": 0,
            "frozen_encoder_input_tokens_including_prompt": 0,
            "scope": "entire invocation, including training controls, profiling and evaluation",
            "includes_rejected_and_profiling_generations": true
          },
          "protocol": {
            "test_opened": false,
            "gold_in_student_generation": false,
            "train_examples": 8,
            "validation_examples": 4,
            "max_tokens": 512
          },
          "parameters_changed": true,
          "variant": "staged",
          "difficulty_history": [
            [
              0.75,
              0.5,
              0.25,
              0.25,
              0,
              0.75,
              0,
              0
            ],
            [
              0.875,
              0.25,
              0,
              0.625,
              0.125,
              0.75,
              0,
              0.125
            ],
            [
              1,
              0.375,
              0.125,
              0.75,
              0.125,
              0.75,
              0,
              0.125
            ],
            [
              0.875,
              0.5,
              0,
              0.625,
              0,
              0.875,
              0.25,
              0
            ]
          ],
          "difficulty_dynamics": {
            "span": [
              0.25,
              0.25,
              0.25,
              0.5,
              0.125,
              0.125,
              0.25,
              0.125
            ],
            "total_variation": [
              0.375,
              0.5,
              0.5,
              0.625,
              0.25,
              0.125,
              0.25,
              0.25
            ],
            "extra_round_trip_variation": [
              0.125,
              0.25,
              0.25,
              0.125,
              0.125,
              0,
              0,
              0.125
            ]
          },
          "initial_static_easy_to_hard_order": null,
          "execution_status": "passed",
          "limitations": [
            "Small public train-only GSM8K holdout diagnostic; no official test is read.",
            "Qwen2.5-0.5B replaces 1.5B and diagnostic cadence is scaled to each bounded step.",
            "Run fixed-medium/staged/too-hard/unfiltered/static separately with matched configured budgets.",
            "All rejected groups and calibration generations are charged; no invented nonzero advantage.",
            "Staged diagnostic uses the validation split, not original 128-per-band training-heldout pools."
          ]
        },
        "too-hard": {
          "method": "grpo-difficulty-curriculum",
          "seed": 42,
          "device": "cuda",
          "diagnostic_only": true,
          "promotion_eligible": false,
          "evaluation_tier": "l1_mechanism",
          "dataset": {
            "verified": true,
            "test_read": false,
            "source_id": "openai/grade-school-math",
            "source_revision": "3101c7d5072418e28b9008a6636bde82a006892c",
            "source_sha256": "17f347dc51477c50d4efb83959dbb7c56297aba886e5544ee2aaed3024813465",
            "partition_sha256": {
              "train": "6b8672439c8cf6f6b6e65bf5f0e515fbf4c497c9066eefb62a6a6a61ced158e3",
              "validation": "2cf1119ad7b2e31e825274487ef32c073e03a8ed44f08ecd3e0366c1f40ebdb7"
            },
            "partition": "normalized-question sha256 unsigned-prefix / 2^64 < 0.1 validation"
          },
          "checkpoint": {
            "public_id": "Qwen/Qwen2.5-0.5B-Instruct",
            "revision": "7ae557604adf67be50417f59c2c2f167def9a775",
            "verified_mirror": {
              "public_id": "Qwen/Qwen2.5-0.5B-Instruct",
              "revision": "7ae557604adf67be50417f59c2c2f167def9a775",
              "file_sha256": {
                "config.json": "18e18afcaccafade98daf13a54092927904649e1dd4eba8299ab717d5d94ff45",
                "generation_config.json": "e558847a8b4402616f1273797b015104dc266fe4b520056fca88823ba8f8ebe6",
                "merges.txt": "599bab54075088774b1733fde865d5bd747cbcc7a547c5bc12610e874e26f5e3",
                "model.safetensors": "fdf756fa7fcbe7404d5c60e26bff1a0c8b8aa1f72ced49e7dd0210fe288fb7fe",
                "tokenizer.json": "c0382117ea329cdf097041132f6d735924b697924d6f6fc3945713e96ce87539",
                "tokenizer_config.json": "5b5d4f65d0acd3b2d56a35b56d374a36cbc1c8fa5cf3b3febbbfabf22f359583",
                "vocab.json": "ca10d7e9fb3ed18575dd1e277a2579c16d108e32f27439684afa0e10b1440910"
              }
            }
          },
          "baseline": {
            "accuracy": 0.5,
            "examples": 4,
            "generated_tokens": 1419,
            "truncated_responses": 1
          },
          "method_metrics": {
            "accuracy": 0.5,
            "examples": 4,
            "generated_tokens": 1487,
            "truncated_responses": 0
          },
          "training": [
            {
              "step": 1,
              "band": [
                0.125,
                0.375
              ],
              "groups": [
                {
                  "training_row": 0,
                  "eligible": false,
                  "current_success": 0.75
                },
                {
                  "training_row": 1,
                  "eligible": false,
                  "current_success": 0.5
                },
                {
                  "training_row": 2,
                  "eligible": true,
                  "current_success": 0.25,
                  "correctness_rewards": [
                    0,
                    0,
                    0,
                    1,
                    0,
                    1,
                    0,
                    0
                  ],
                  "loss": -1.4901161193847656e-08,
                  "gradient_l2_norm": 0.14908787608146667,
                  "optimizer_update_executed": true,
                  "zero_advantage_group": false,
                  "rewards": [
                    0.2,
                    0.2,
                    0.2,
                    1.2,
                    0.2,
                    1.2,
                    0.2,
                    0.2
                  ],
                  "advantages": [
                    -0.5773489475250244,
                    -0.5773489475250244,
                    -0.5773489475250244,
                    1.7320468425750732,
                    -0.5773489475250244,
                    1.7320468425750732,
                    -0.5773489475250244,
                    -0.5773489475250244
                  ]
                },
                {
                  "training_row": 3,
                  "eligible": true,
                  "current_success": 0.25,
                  "correctness_rewards": [
                    0,
                    1,
                    1,
                    0,
                    0,
                    0,
                    0,
                    0
                  ],
                  "loss": 0.0004897117614746094,
                  "gradient_l2_norm": 0.17364460229873657,
                  "optimizer_update_executed": true,
                  "zero_advantage_group": false,
                  "rewards": [
                    0.2,
                    1.2,
                    1.2,
                    0.2,
                    0.2,
                    0.2,
                    0.2,
                    0.2
                  ],
                  "advantages": [
                    -0.5773490071296692,
                    1.7320467233657837,
                    1.7320467233657837,
                    -0.5773490071296692,
                    -0.5773490071296692,
                    -0.5773490071296692,
                    -0.5773490071296692,
                    -0.5773490071296692
                  ]
                },
                {
                  "training_row": 4,
                  "eligible": false,
                  "current_success": 0
                },
                {
                  "training_row": 5,
                  "eligible": false,
                  "current_success": 0.75
                },
                {
                  "training_row": 6,
                  "eligible": false,
                  "current_success": 0
                },
                {
                  "training_row": 7,
                  "eligible": false,
                  "current_success": 0
                }
              ],
              "validation_diagnostic": {
                "accuracy": 0.5,
                "examples": 4,
                "generated_tokens": 1497,
                "truncated_responses": 0
              },
              "transition": "not-staged",
              "effective_updates": 2
            },
            {
              "step": 2,
              "band": [
                0.125,
                0.375
              ],
              "groups": [
                {
                  "training_row": 0,
                  "eligible": false,
                  "current_success": 0.75
                },
                {
                  "training_row": 1,
                  "eligible": true,
                  "current_success": 0.125,
                  "correctness_rewards": [
                    1,
                    0,
                    0,
                    0,
                    0,
                    0,
                    0,
                    0
                  ],
                  "loss": 1.3783574104309082e-07,
                  "gradient_l2_norm": 0.15373632311820984,
                  "optimizer_update_executed": true,
                  "zero_advantage_group": false,
                  "rewards": [
                    1.2,
                    0.2,
                    0.2,
                    0.2,
                    0.2,
                    0.2,
                    0.2,
                    0.2
                  ],
                  "advantages": [
                    2.6457431316375732,
                    -0.37796342372894287,
                    -0.37796342372894287,
                    -0.37796342372894287,
                    -0.37796342372894287,
                    -0.37796342372894287,
                    -0.37796342372894287,
                    -0.37796342372894287
                  ]
                },
                {
                  "training_row": 2,
                  "eligible": true,
                  "current_success": 0.375,
                  "correctness_rewards": [
                    0,
                    0,
                    1,
                    1,
                    0,
                    1,
                    0,
                    0
                  ],
                  "loss": 0.0007002353668212891,
                  "gradient_l2_norm": 0.16592727601528168,
                  "optimizer_update_executed": true,
                  "zero_advantage_group": false,
                  "rewards": [
                    0.2,
                    0.2,
                    1.2,
                    1.2,
                    0.2,
                    1.2,
                    0.2,
                    0.2
                  ],
                  "advantages": [
                    -0.7745950222015381,
                    -0.7745950222015381,
                    1.2909917831420898,
                    1.2909917831420898,
                    -0.7745950222015381,
                    1.2909917831420898,
                    -0.7745950222015381,
                    -0.7745950222015381
                  ]
                },
                {
                  "training_row": 3,
                  "eligible": false,
                  "current_success": 0.5
                },
                {
                  "training_row": 4,
                  "eligible": true,
                  "current_success": 0.125,
                  "correctness_rewards": [
                    0,
                    1,
                    0,
                    0,
                    0,
                    0,
                    0,
                    0
                  ],
                  "loss": 0.00017904490232467651,
                  "gradient_l2_norm": 0.1151842400431633,
                  "optimizer_update_executed": true,
                  "zero_advantage_group": false,
                  "rewards": [
                    0.2,
                    1.2,
                    0.2,
                    0.2,
                    0.2,
                    0.2,
                    0.2,
                    0.2
                  ],
                  "advantages": [
                    -0.37796345353126526,
                    2.6457433700561523,
                    -0.37796345353126526,
                    -0.37796345353126526,
                    -0.37796345353126526,
                    -0.37796345353126526,
                    -0.37796345353126526,
                    -0.37796345353126526
                  ]
                },
                {
                  "training_row": 5,
                  "eligible": false,
                  "current_success": 0.75
                },
                {
                  "training_row": 6,
                  "eligible": false,
                  "current_success": 0
                },
                {
                  "training_row": 7,
                  "eligible": false,
                  "current_success": 0
                }
              ],
              "validation_diagnostic": {
                "accuracy": 0.5,
                "examples": 4,
                "generated_tokens": 1650,
                "truncated_responses": 1
              },
              "transition": "not-staged",
              "effective_updates": 3
            },
            {
              "step": 3,
              "band": [
                0.125,
                0.375
              ],
              "groups": [
                {
                  "training_row": 0,
                  "eligible": false,
                  "current_success": 1
                },
                {
                  "training_row": 1,
                  "eligible": true,
                  "current_success": 0.25,
                  "correctness_rewards": [
                    1,
                    0,
                    0,
                    0,
                    0,
                    0,
                    1,
                    0
                  ],
                  "loss": 8.195638656616211e-08,
                  "gradient_l2_norm": 0.15684774518013,
                  "optimizer_update_executed": true,
                  "zero_advantage_group": false,
                  "rewards": [
                    1.2,
                    0.2,
                    0.2,
                    0.2,
                    0.2,
                    0.2,
                    1.2,
                    0.2
                  ],
                  "advantages": [
                    1.7320467233657837,
                    -0.5773490071296692,
                    -0.5773490071296692,
                    -0.5773490071296692,
                    -0.5773490071296692,
                    -0.5773490071296692,
                    1.7320467233657837,
                    -0.5773490071296692
                  ]
                },
                {
                  "training_row": 2,
                  "eligible": false,
                  "current_success": 0.5
                },
                {
                  "training_row": 3,
                  "eligible": false,
                  "current_success": 0.5
                },
                {
                  "training_row": 4,
                  "eligible": false,
                  "current_success": 0
                },
                {
                  "training_row": 5,
                  "eligible": false,
                  "current_success": 1
                },
                {
                  "training_row": 6,
                  "eligible": true,
                  "current_success": 0.25,
                  "correctness_rewards": [
                    0,
                    0,
                    0,
                    1,
                    0,
                    1,
                    0,
                    0
                  ],
                  "loss": -0.0002179965376853943,
                  "gradient_l2_norm": 0.16200219094753265,
                  "optimizer_update_executed": true,
                  "zero_advantage_group": false,
                  "rewards": [
                    0.2,
                    0.2,
                    0.2,
                    1.2,
                    0.2,
                    1.2,
                    0.2,
                    0.2
                  ],
                  "advantages": [
                    -0.5773489475250244,
                    -0.5773489475250244,
                    -0.5773489475250244,
                    1.7320468425750732,
                    -0.5773489475250244,
                    1.7320468425750732,
                    -0.5773489475250244,
                    -0.5773489475250244
                  ]
                },
                {
                  "training_row": 7,
                  "eligible": false,
                  "current_success": 0
                }
              ],
              "validation_diagnostic": {
                "accuracy": 0.75,
                "examples": 4,
                "generated_tokens": 1240,
                "truncated_responses": 0
              },
              "transition": "not-staged",
              "effective_updates": 2
            },
            {
              "step": 4,
              "band": [
                0.125,
                0.375
              ],
              "groups": [
                {
                  "training_row": 0,
                  "eligible": false,
                  "current_success": 0.625
                },
                {
                  "training_row": 1,
                  "eligible": false,
                  "current_success": 0.625
                },
                {
                  "training_row": 2,
                  "eligible": true,
                  "current_success": 0.375,
                  "correctness_rewards": [
                    1,
                    0,
                    1,
                    1,
                    0,
                    0,
                    0,
                    0
                  ],
                  "loss": -1.862645149230957e-07,
                  "gradient_l2_norm": 0.17808710038661957,
                  "optimizer_update_executed": true,
                  "zero_advantage_group": false,
                  "rewards": [
                    1.2,
                    0.2,
                    1.2,
                    1.2,
                    0.2,
                    0.2,
                    0.2,
                    0.2
                  ],
                  "advantages": [
                    1.2909919023513794,
                    -0.7745948433876038,
                    1.2909919023513794,
                    1.2909919023513794,
                    -0.7745948433876038,
                    -0.7745948433876038,
                    -0.7745948433876038,
                    -0.7745948433876038
                  ]
                },
                {
                  "training_row": 3,
                  "eligible": true,
                  "current_success": 0.375,
                  "correctness_rewards": [
                    0,
                    1,
                    0,
                    0,
                    1,
                    0,
                    0,
                    1
                  ],
                  "loss": -0.0007097646594047546,
                  "gradient_l2_norm": 0.16549642384052277,
                  "optimizer_update_executed": true,
                  "zero_advantage_group": false,
                  "rewards": [
                    0.2,
                    1.2,
                    0.2,
                    0.2,
                    1.2,
                    0.2,
                    0.2,
                    1.2
                  ],
                  "advantages": [
                    -0.7745951414108276,
                    1.2909916639328003,
                    -0.7745951414108276,
                    -0.7745951414108276,
                    1.2909916639328003,
                    -0.7745951414108276,
                    -0.7745951414108276,
                    1.2909916639328003
                  ]
                },
                {
                  "training_row": 4,
                  "eligible": true,
                  "current_success": 0.125,
                  "correctness_rewards": [
                    0,
                    0,
                    1,
                    0,
                    0,
                    0,
                    0,
                    0
                  ],
                  "loss": 0.0005903393030166626,
                  "gradient_l2_norm": 0.14686866104602814,
                  "optimizer_update_executed": true,
                  "zero_advantage_group": false,
                  "rewards": [
                    0.2,
                    0.2,
                    1.2,
                    0.2,
                    0.2,
                    0.2,
                    0.2,
                    0.2
                  ],
                  "advantages": [
                    -0.3779633343219757,
                    -0.3779633343219757,
                    2.6457431316375732,
                    -0.3779633343219757,
                    -0.3779633343219757,
                    -0.3779633343219757,
                    -0.3779633343219757,
                    -0.3779633343219757
                  ]
                },
                {
                  "training_row": 5,
                  "eligible": false,
                  "current_success": 0.75
                },
                {
                  "training_row": 6,
                  "eligible": false,
                  "current_success": 0
                },
                {
                  "training_row": 7,
                  "eligible": true,
                  "current_success": 0.125,
                  "correctness_rewards": [
                    0,
                    0,
                    0,
                    0,
                    0,
                    0,
                    1,
                    0
                  ],
                  "loss": -0.00016710534691810608,
                  "gradient_l2_norm": 0.17707861959934235,
                  "optimizer_update_executed": true,
                  "zero_advantage_group": false,
                  "rewards": [
                    0.2,
                    0.2,
                    0.2,
                    0.2,
                    0.2,
                    0.2,
                    1.2,
                    0.2
                  ],
                  "advantages": [
                    -0.3779633343219757,
                    -0.3779633343219757,
                    -0.3779633343219757,
                    -0.3779633343219757,
                    -0.3779633343219757,
                    -0.3779633343219757,
                    2.6457431316375732,
                    -0.3779633343219757
                  ]
                }
              ],
              "validation_diagnostic": {
                "accuracy": 0.5,
                "examples": 4,
                "generated_tokens": 1487,
                "truncated_responses": 0
              },
              "transition": "not-staged",
              "effective_updates": 4
            }
          ],
          "cost": {
            "generation_calls": {
              "evaluation": 24,
              "student": 32
            },
            "generated_tokens": {
              "evaluation": 8780,
              "student": 74351
            },
            "scored_response_tokens": {
              "behavior": 74351,
              "student_rl": 26978
            },
            "frozen_encoder_forward_calls": 0,
            "frozen_encoder_input_tokens_including_prompt": 0,
            "scope": "entire invocation, including training controls, profiling and evaluation",
            "includes_rejected_and_profiling_generations": true
          },
          "protocol": {
            "test_opened": false,
            "gold_in_student_generation": false,
            "train_examples": 8,
            "validation_examples": 4,
            "max_tokens": 512
          },
          "parameters_changed": true,
          "variant": "too-hard",
          "difficulty_history": [
            [
              0.75,
              0.5,
              0.25,
              0.25,
              0,
              0.75,
              0,
              0
            ],
            [
              0.75,
              0.125,
              0.375,
              0.5,
              0.125,
              0.75,
              0,
              0
            ],
            [
              1,
              0.25,
              0.5,
              0.5,
              0,
              1,
              0.25,
              0
            ],
            [
              0.625,
              0.625,
              0.375,
              0.375,
              0.125,
              0.75,
              0,
              0.125
            ]
          ],
          "difficulty_dynamics": {
            "span": [
              0.375,
              0.5,
              0.25,
              0.25,
              0.125,
              0.25,
              0.25,
              0.125
            ],
            "total_variation": [
              0.625,
              0.875,
              0.375,
              0.375,
              0.375,
              0.5,
              0.5,
              0.125
            ],
            "extra_round_trip_variation": [
              0.25,
              0.375,
              0.125,
              0.125,
              0.25,
              0.25,
              0.25,
              0
            ]
          },
          "initial_static_easy_to_hard_order": null,
          "execution_status": "passed",
          "limitations": [
            "Small public train-only GSM8K holdout diagnostic; no official test is read.",
            "Qwen2.5-0.5B replaces 1.5B and diagnostic cadence is scaled to each bounded step.",
            "Run fixed-medium/staged/too-hard/unfiltered/static separately with matched configured budgets.",
            "All rejected groups and calibration generations are charged; no invented nonzero advantage.",
            "Staged diagnostic uses the validation split, not original 128-per-band training-heldout pools."
          ]
        }
      }
    }
  ],
  "aggregate_metrics": {
    "fixed-medium.initial.accuracy": {
      "mean": 0.5,
      "std": 0,
      "n": 1
    },
    "fixed-medium.validation.accuracy": {
      "mean": 0.75,
      "std": 0,
      "n": 1
    },
    "staged.initial.accuracy": {
      "mean": 0.5,
      "std": 0,
      "n": 1
    },
    "staged.validation.accuracy": {
      "mean": 0.5,
      "std": 0,
      "n": 1
    },
    "too-hard.initial.accuracy": {
      "mean": 0.5,
      "std": 0,
      "n": 1
    },
    "too-hard.validation.accuracy": {
      "mean": 0.5,
      "std": 0,
      "n": 1
    }
  },
  "execution_checks": {
    "fixed-medium": {
      "actual_optimizer_updates": 8,
      "invocation_elapsed_seconds": 538.1836995203048,
      "validation_examples": 4,
      "difficulty_recalibration_steps": 4,
      "actual_bands": [
        [
          0.375,
          0.625
        ],
        [
          0.375,
          0.625
        ],
        [
          0.375,
          0.625
        ],
        [
          0.375,
          0.625
        ]
      ],
      "actual_stage_transitions": [
        "not-staged",
        "not-staged",
        "not-staged",
        "not-staged"
      ],
      "measured_cost": {
        "generation_calls": {
          "evaluation": 24,
          "student": 32
        },
        "generated_tokens": {
          "evaluation": 8413,
          "student": 72642
        },
        "scored_response_tokens": {
          "behavior": 72642,
          "student_rl": 18142
        },
        "frozen_encoder_forward_calls": 0,
        "frozen_encoder_input_tokens_including_prompt": 0,
        "scope": "entire invocation, including training controls, profiling and evaluation",
        "includes_rejected_and_profiling_generations": true
      }
    },
    "staged": {
      "actual_optimizer_updates": 9,
      "invocation_elapsed_seconds": 562.0409296359867,
      "validation_examples": 4,
      "difficulty_recalibration_steps": 4,
      "actual_bands": [
        [
          0.625,
          0.875
        ],
        [
          0.625,
          0.875
        ],
        [
          0.625,
          0.875
        ],
        [
          0.5,
          0.75
        ]
      ],
      "actual_stage_transitions": [
        "new-best",
        "keep",
        "advance",
        "new-best"
      ],
      "measured_cost": {
        "generation_calls": {
          "evaluation": 24,
          "student": 32
        },
        "generated_tokens": {
          "evaluation": 9158,
          "student": 73273
        },
        "scored_response_tokens": {
          "behavior": 73273,
          "student_rl": 18844
        },
        "frozen_encoder_forward_calls": 0,
        "frozen_encoder_input_tokens_including_prompt": 0,
        "scope": "entire invocation, including training controls, profiling and evaluation",
        "includes_rejected_and_profiling_generations": true
      }
    },
    "too-hard": {
      "actual_optimizer_updates": 11,
      "invocation_elapsed_seconds": 555.9015793707222,
      "validation_examples": 4,
      "difficulty_recalibration_steps": 4,
      "actual_bands": [
        [
          0.125,
          0.375
        ],
        [
          0.125,
          0.375
        ],
        [
          0.125,
          0.375
        ],
        [
          0.125,
          0.375
        ]
      ],
      "actual_stage_transitions": [
        "not-staged",
        "not-staged",
        "not-staged",
        "not-staged"
      ],
      "measured_cost": {
        "generation_calls": {
          "evaluation": 24,
          "student": 32
        },
        "generated_tokens": {
          "evaluation": 8780,
          "student": 74351
        },
        "scored_response_tokens": {
          "behavior": 74351,
          "student_rl": 26978
        },
        "frozen_encoder_forward_calls": 0,
        "frozen_encoder_input_tokens_including_prompt": 0,
        "scope": "entire invocation, including training controls, profiling and evaluation",
        "includes_rejected_and_profiling_generations": true
      }
    }
  },
  "actual_cuda_run_count": 3,
  "dataset": {
    "verified": true,
    "test_read": false,
    "source_id": "openai/grade-school-math",
    "source_revision": "3101c7d5072418e28b9008a6636bde82a006892c",
    "source_sha256": "17f347dc51477c50d4efb83959dbb7c56297aba886e5544ee2aaed3024813465",
    "partition_sha256": {
      "train": "6b8672439c8cf6f6b6e65bf5f0e515fbf4c497c9066eefb62a6a6a61ced158e3",
      "validation": "2cf1119ad7b2e31e825274487ef32c073e03a8ed44f08ecd3e0366c1f40ebdb7"
    },
    "partition": "normalized-question sha256 unsigned-prefix / 2^64 < 0.1 validation"
  },
  "checkpoint": {
    "public_id": "Qwen/Qwen2.5-0.5B-Instruct",
    "revision": "7ae557604adf67be50417f59c2c2f167def9a775",
    "verified_mirror": {
      "public_id": "Qwen/Qwen2.5-0.5B-Instruct",
      "revision": "7ae557604adf67be50417f59c2c2f167def9a775",
      "file_sha256": {
        "config.json": "18e18afcaccafade98daf13a54092927904649e1dd4eba8299ab717d5d94ff45",
        "generation_config.json": "e558847a8b4402616f1273797b015104dc266fe4b520056fca88823ba8f8ebe6",
        "merges.txt": "599bab54075088774b1733fde865d5bd747cbcc7a547c5bc12610e874e26f5e3",
        "model.safetensors": "fdf756fa7fcbe7404d5c60e26bff1a0c8b8aa1f72ced49e7dd0210fe288fb7fe",
        "tokenizer.json": "c0382117ea329cdf097041132f6d735924b697924d6f6fc3945713e96ce87539",
        "tokenizer_config.json": "5b5d4f65d0acd3b2d56a35b56d374a36cbc1c8fa5cf3b3febbbfabf22f359583",
        "vocab.json": "ca10d7e9fb3ed18575dd1e277a2579c16d108e32f27439684afa0e10b1440910"
      }
    }
  },
  "evaluation_protocol": {
    "tier": "l1_mechanism",
    "seeds": [
      42
    ],
    "formal_comparison": false,
    "promotion_eligible": false,
    "test_used_for_selection": false,
    "matched_configured_budget": true,
    "matched_actual_compute": false,
    "claim_policy": "single-seed mechanism smoke, not stable capability or production improvement"
  },
  "provenance": {
    "code_commit": "6aabe57a97ec3a586aac7aeea66128750f9441da",
    "artifact_path": "docs/reproductions/2610.13669-grpo-difficulty-curriculum/metrics/public-a100-three-arms-seed42.json",
    "dataset_fingerprint": "69efc634f51a589335fcdea85238bf3218653a94de2ba2604a64056960547dab",
    "execution_config_sha256": "506dce6568921b9c8fa75d969267537f9bb5c5c5b093a77b283616a2246524c5",
    "raw_results_sha256": {
      "fixed-medium": "49b6015c9161672607353e73c8c9d8d0625c1da2354ffd29a36a2694787fc469",
      "staged": "26b59149c4c36cf60cc9ca900042f82ebe37cf9c96c5b40f49bcafaeda9740e0",
      "too-hard": "00242a21ba68dc59e5468af3ae570ac715563290756141f50bd8e8b89465bf46"
    }
  },
  "limitations": [
    "One seed and three variants are NOT three independent seeds; L1 mechanism smoke only.",
    "Four validation examples cannot establish stable gains or original B5 degradation; test stayed isolated.",
    "Matched configured rollout caps do not imply equal effective updates, realized token cost or runtime.",
    "Only fixed-medium/staged/B5-too-hard were executed; static and unfiltered controls are not claimed.",
    "Original default seeds 43/44 remain unfinished; passing this CUDA smoke does not complete the batch plan.",
    "Staged reached B2, not the full original B1-to-B5 curriculum; validation cadence is reduced-scale."
  ]
}