🔬Modal Research Exchange

{
  "activation_gate": {
    "applicable": true,
    "checks": [
      {
        "max": 0.3,
        "measures": "fraction of loss tokens masked by the DPPO divergence rule on the second minibatch update, mean over pilot steps",
        "metric": "dppo_masked_token_fraction",
        "min": 0.001
      },
      {
        "measures": "90th percentile of |p_current - p_sampler| over loss tokens at the second update",
        "metric": "policy_drift_abs_p_q90",
        "min": 0.005
      },
      {
        "measures": "fraction of tokens clipped by the eps=0.2 ratio clip in the baseline loss on the same pilot batches (confirms the regime is off-policy for the baseline too)",
        "metric": "grpo_clip_fraction",
        "min": 0.0005
      },
      {
        "max": 2,
        "measures": "optim_step calls per rollout batch as logged",
        "metric": "executed_updates_per_batch",
        "min": 2
      },
      {
        "measures": "model name reported at client creation",
        "metric": "executed_base_model",
        "one_of": [
          "Qwen/Qwen3.5-2B"
        ]
      }
    ],
    "reason": "DPPO only differs from GRPO on tokens whose divergence exceeds the trust region. A 5-step pilot of the treatment arm, with 2 updates per batch, must show that the mask fires on a nontrivial but not dominant share of tokens and that the policy actually drifts. Executed config values must match the plan."
  },
  "arms": [
    {
      "loss": "grpo_clip_eps0.2",
      "lr": 2e-05,
      "name": "grpo_seed0",
      "seed": 0,
      "updates_per_batch": 2
    },
    {
      "loss": "dppo_divergence_mask",
      "lr": 2e-05,
      "name": "dppo_seed0",
      "seed": 0,
      "updates_per_batch": 2
    }
  ],
  "base_model": "Qwen/Qwen3.5-2B",
  "baseline": "GRPO clipped-ratio loss (eps=0.2) through the same custom loss path. Identical rollouts per step, 2 minibatch updates per batch, lr, rank, seed, data order and eval.",
  "batch_size": 256,
  "budget": {
    "gpu_hours": 2.0,
    "max_wall_minutes": 120
  },
  "command": "Custom tinker SDK loop; the loss function is the custom part. Sketch:\n\nimport os, tinker\nsc = tinker.ServiceClient(base_url=os.environ['TINKER_BASE_URL'])\ntc = sc.create_lora_training_client(base_model='Qwen/Qwen3.5-2B', rank=32)\n\n# Rollouts: reuse tinker_cookbook.recipes.math_rl GSM8K env and renderer.\n# group_size=8, groups_per_batch=32, max_tokens=512, seed=0.\nfor step in range(50):\n    sampler = tc.save_weights_and_get_sampling_client(f's{step}')\n    batch = rollout(sampler, gsm8k_train_prompts[step], group_size=8)  # stores sampler logprobs as old_lp\n    adv = group_mean_center(batch.rewards)\n    for mb in split(batch, 2):\n        tc.forward_backward_custom(mb.data, make_loss(LOSS, old_lp=mb.old_lp, adv=mb.adv)).result()\n        tc.optim_step(tinker.AdamParams(learning_rate=2e-5)).result()\n\n# LOSS='grpo': -min(r*A, clip(r, 0.8, 1.2)*A)\n# LOSS='dppo': -(r*A) * keep_mask(div(p, p_old) <= delta), with div and delta pinned from the paper code\n# Final eval: greedy decoding on openai/gsm8k test (1319 items), numeric exact match.",
  "correctness_preflight": {
    "paths": [
      {
        "baseline_revision": "UNRESOLVED: pin tinker-cookbook commit whose stock ppo/importance-sampling loss defines GRPO clipping",
        "id": "grpo_dppo_custom_loss",
        "kind": "loss",
        "reference_revision": "UNRESOLVED: pin commit of the official DPPO implementation for arXiv 2602.04879",
        "tolerances": {
          "baseline_gradient": {
            "atol": 1e-05,
            "min_cosine": 0.9999,
            "rtol": 0.001
          },
          "baseline_loss": {
            "atol": 1e-05,
            "rtol": 0.0001
          },
          "reference_gradient": {
            "atol": 1e-05,
            "min_cosine": 0.9999,
            "rtol": 0.001
          },
          "reference_loss": {
            "atol": 1e-05,
            "rtol": 0.0001
          }
        }
      }
    ],
    "required": true,
    "status": "unresolved: DPPO reference code revision and exact divergence/threshold not yet pinned (paper digest failed); no full sweep until resolved"
  },
  "dataset": {
    "hf_path": "openai/gsm8k",
    "n_examples": 1600,
    "name": "gsm8k",
    "split": "main/train"
  },
  "decision_rule": {
    "alpha": 0.05,
    "criteria": [
      {
        "bounds": [
          0,
          1
        ],
        "id": "primary",
        "margin": 0.0,
        "metric": "accuracy",
        "weights": {
          "baseline": -1,
          "treatment": 1
        }
      }
    ],
    "estimand": "Difference in greedy GSM8K test exact-match accuracy, DPPO minus GRPO, for these fixed final checkpoints of Qwen3.5-2B LoRA; larger is better",
    "method": "paired_hoeffding",
    "seeds": [
      0
    ],
    "unit": "fixed_checkpoint_items",
    "version": 1
  },
  "eval": {
    "aggregation": null,
    "benchmark": "GSM8K test (openai/gsm8k main/test)",
    "metric": "greedy exact-match accuracy",
    "n_examples": 1319
  },
  "experiment_features": {
    "adaptive_selection": false,
    "checkpoint_restore": false,
    "custom_loss": true,
    "custom_optimizer": false,
    "generated_training_data": false,
    "heldout_loss": false,
    "learned_generator": false,
    "paired_comparison": true,
    "paired_training_arms": true,
    "warmup_adaptation": false
  },
  "feasibility": null,
  "feasibility_check": {
    "gpu_hours": null,
    "known_gpu_hours": 0,
    "phases": {},
    "problems": [
      "Measured feasibility not recorded"
    ],
    "status": "unknown",
    "wall_seconds": null
  },
  "feasibility_required": true,
  "generation_audit": {
    "answer_comparison": "numeric_exact",
    "max_eval_overlaps": 0,
    "max_label_errors": 0,
    "min_sample_size": 100,
    "required": false
  },
  "hypothesis": "On Qwen3.5-2B LoRA RL on GSM8K with 2 policy updates per rollout batch, the DPPO loss reaches higher greedy exact-match accuracy on the 1319-item GSM8K test set than the GRPO clipped-ratio loss. The lower bound of the paired 95% CI on DPPO minus GRPO must be above 0.",
  "lora_rank": 32,
  "max_seq_len": 1024,
  "method": "The requested Qwen3.5-0.8B is not served, so we use the closest served small model, Qwen/Qwen3.5-2B.\n\nBoth arms run the same client-side RL loop: the tinker-cookbook math_rl GSM8K env, renderer, prompts, reward and data order. Each step samples groups_per_batch=32 GSM8K train prompts with group_size=8 rollouts at temperature 1.0 and max_tokens=512. Advantages are group-mean-centered (GRPO).\n\nThe rollout batch is split into 2 minibatches. That gives 2 optim_steps per batch, so the policy drifts from the sampling policy and the trust region can act. With one update per fresh batch, neither mask nor clip fires. Both losses are computed in one forked custom loss path through forward_backward_custom, using the sampler logprobs as p_old.\n\nThe arms differ in exactly one switch:\n- Baseline (GRPO): token loss -min(r*A, clip(r, 1-0.2, 1+0.2)*A).\n- Treatment (DPPO, arXiv 2602.04879): token loss -r*A, with tokens masked out when the divergence between the current and sampling policy exceeds the threshold delta in the direction the advantage pushes. The divergence is estimated from the token probabilities, e.g. |p - p_old|, rather than the single-sample ratio.\n\nTriage only received a truncated digest of the paper. The exact divergence estimator, delta and masking direction must be pinned from the paper's released code before any training. Until then the correctness preflight is unresolved and no full sweep runs.\n\nTraining: LoRA rank 32, AdamW lr 2e-5, 50 rollout steps (100 optim steps) per arm, seed 0. Both arms start from the same base weights.\n\nEvaluation: the final checkpoint of each arm is scored with greedy decoding on all 1319 GSM8K test items, using the cookbook's numeric answer grader.",
  "optimizer": {
    "lr": 2e-05,
    "name": "adam"
  },
  "primary_endpoint": "gsm8k_test_greedy_exact_match_accuracy@final_checkpoint (1319 items)",
  "restoration_preflight": {
    "applicable": true,
    "fixed_batch_required": true,
    "groups": [
      {
        "arms": [
          {
            "name": "grpo_seed0",
            "optimizer": "adam",
            "optimizer_config": {
              "beta1": 0.9,
              "beta2": 0.95,
              "eps": 1e-08,
              "learning_rate": 2e-05
            }
          },
          {
            "name": "dppo_seed0",
            "optimizer": "adam",
            "optimizer_config": {
              "beta1": 0.9,
              "beta2": 0.95,
              "eps": 1e-08,
              "learning_rate": 2e-05
            }
          }
        ]
      }
    ],
    "reason": "Both arms are LoRA-trained from the same base. Before the first update, verify identical initial state: compare the logprobs on a fixed tokenized GSM8K batch across both freshly created clients, which must be bitwise or within 1e-6. Also verify the active optimizer is AdamW with the planned config after creation. Server identity cannot be introspected and stays unverified."
  },
  "steps": 100,
  "success_criterion": "On the 1319 GSM8K test items, the paired 95% Hoeffding CI for DPPO accuracy minus GRPO accuracy has a lower bound above 0.0, roughly a 7.5-point gap at this n. A smaller positive gap is reported as inconclusive, not validated.",
  "title": "DPPO divergence-masked loss vs GRPO ratio-clipped loss, Qwen3.5-2B GSM8K LoRA RL",
  "training_type": "rl"
}

[
  {
    "arm": "dppo_diag",
    "step": 16,
    "uri": "tinker://daecfa24-b646-4b82-b337-4447b0e95af0:train:0/weights/rx21b_diag_final"
  },
  {
    "arm": "dppo_diag",
    "step": 0,
    "uri": "tinker://21e32976f18f971463410adff4e364e4/weights/rx19badd842c_init"
  }
]

{
  "dppo_diag": {
    "epoch2_abs_drift_gt_delta_tokens": 224,
    "epoch2_dppo_masked_nonzero_adv_fraction": 0.0012842981177005712,
    "epoch2_dppo_masked_token_fraction": 0.0011361484567316796,
    "epoch2_n_loss_tokens": 84496,
    "epoch2_policy_drift_abs_p_max": 0.4187354204836845,
    "epoch2_policy_drift_abs_p_q90": 0.023858560402555296,
    "epoch2_policy_drift_abs_p_q99": 0.08628262678867459,
    "epoch2_ppo_counterfactual_clip_fraction": 0.013692955879568263,
    "epoch2_ratio_cap_hits": 0,
    "epoch2_removed_nonzero_grad_neg": 46,
    "epoch2_removed_nonzero_grad_pos": 50
  }
}

diag_updates_executed

16

diag_wall_seconds

470.8

epoch1_dppo_masked_token_fraction

0

epoch1_ppo_counterfactual_clip_fraction

0.001195

epoch1_policy_drift_abs_p_q90

0.01251

epoch1_policy_drift_abs_p_max

0.137

sentinel_init_vs_sampler_absdp_q90

0.007642

sentinel_init_vs_sampler_absdp_max

0.06218

sampled_reward_mean

0.5789

executed_updates_per_batch

16

sentinel_after_u1_vs_init_absdp_q90

0.005858

sentinel_after_u1_vs_init_absdp_max

0.06052

sentinel_after_u2_vs_init_absdp_q90

0.007421

sentinel_after_u2_vs_init_absdp_max

0.06134

sentinel_after_u4_vs_init_absdp_q90

0.009359

sentinel_after_u4_vs_init_absdp_max

0.09145

sentinel_after_u8_vs_init_absdp_q90

0.01559

sentinel_after_u8_vs_init_absdp_max

0.1537

sentinel_after_u16_vs_init_absdp_q90

0.03011

sentinel_after_u16_vs_init_absdp_max

0.3284

• checkpoint · dppo_diag · step 16 · tinker://daecfa24-b646-4b82-b337-4447b0e95af0:train:0/weights/rx21b_diag_final
• checkpoint · dppo_diag · step 0 · tinker://21e32976f18f971463410adff4e364e4/weights/rx19badd842c_init
• instructions · launch-instructions.md sha256 b42946cbc224
• script · diag.py sha256 1360d041a065
• script · analyze.py sha256 e444bfe6238f
• script · losses_patched.py sha256 bceaad747abb
• script · audit_dppo_evidence.py sha256 f90c441284ff
• code_patch · losses-instrumentation.patch sha256 d558bd82075a
• effective_config · revised-diagnostic-spec.json sha256 7af592135f9b
• effective_config · effective-config.json sha256 7d198e6df468
• effective_config · next-comparison-spec.json sha256 963bda2f5972
• code_revision · code-revision.json sha256 3b77d75e15ef
• metrics · diag-events.jsonl sha256 c9ebbb71d393
• metrics · diag-per-update.json sha256 5c7ac8aa254e
• metrics · diag-analysis.json sha256 be7f0bc4e86b
• metrics · diag-frozen-batch.json sha256 d275384bc777
• metrics · audit-result-original-rerun.json sha256 58933b7bc645
• metrics · audit-result-patched-rerun.json sha256 059303ae2016
• metrics · prior-evidence-manifest.json sha256 ba862f882327
• metrics · prior-pilot-dppo-steps.jsonl sha256 ee68c6b8951c
• metrics · diag-report.md sha256 6576232559a1
• metrics · rx21b-diag-tensors.tar.gz sha256 08735d655a54

1. Nov 3, 22:53 UTC · succeeded · done: dppo_diag · Bounded diagnostic on shared Spindle (Qwen3.5-2B, 16 updates over one frozen 32x8 GSM8K-train batch, 8 minibatches x 2 epochs, 471 s): DPPO mask was 0 in epoch 1 and 0.00114 pooled in epoch 2 (96 rem…
2. Nov 3, 22:40 UTC · running · Bounded DPPO activation diagnostic (requester-authorized continuation of exp_e6b0965af3): one frozen 32x8 GSM8K train batch, 8 minibatches x 2 epochs, <=16 updates, 12-minute cap. Offline replay of t…

correctness_preflight

{'required': True, 'paths': [{'id': 'grpo_dppo_custom_loss', 'status_note': 'Spec revisions are UNRESOLVED placeholders; real pins below. Offline replay only (CPU torch 2.15.0, saved fixture); no new tolerance chosen.', 'fixed_input': {'sha256': 'aff0d4c84ff104219bf365a13dcb0de1c1cfc59b833ac72cb357c46e29a83f13', 'description': 'run_19badd842c preflight_fixture.json (1024 loss tokens, synthetic stale)'}, 'baseline_revision': 'UNRESOLVED: pin tinker-cookbook commit whose stock ppo/importance-sampling loss defines GRPO clipping', 'reference_revision': 'UNRESOLVED: pin commit of the official DPPO implementation for arXiv 2602.04879', 'pinned_baseline': 'modal-projects/spindle@e58ebef130b4cd02de6880e62d27ec358ff7132c forward_backward.py::_loss', 'pinned_reference': 'sail-sg/Stable-RL@c6af1d119c0fe30d16dd9a837d35fdf695acd226 core_algos.compute_policy_loss_with_mask(dppo_binary_tv)', 'implementation_revision': 'losses_patched.py sha256 bceaad747abb77f986d71bdbc8e902d0a0fa3d77bd827401f731a4e663d00a6c (token_terms identical to 36e74485...)', 'precision': 'float32 loss path, CPU', 'normalization': 'sum over masked loss tokens (batch_num_tokens=1, dp_size=1)', 'errors': {'baseline_loss': {'max_abs_error': 0.0, 'reference_max_abs': 429.46087646484375, 'relative_l2_error': 0.0, 'cosine_similarity': 1.0}, 'baseline_gradient': {'max_abs_error': 0.0, 'reference_max_abs': None, 'relative_l2_error': 0.0, 'cosine_similarity': 1.0}, 'reference_loss': {'max_abs_error': 3.0517578125e-05, 'reference_max_abs': 420.54302978515625, 'relative_l2_error': 7.256e-08, 'cosine_similarity': 1.0}, 'reference_gradient': {'max_abs_error': 0.0, 'reference_max_abs': None, 'relative_l2_error': 0.0, 'cosine_similarity': 1.0}}, 'lr_scaling': []}]}

correctness_validation

{'status': 'unverified', 'required': True, 'checks': [{'path': 'grpo_dppo_custom_loss', 'check': 'baseline_loss', 'status': 'passed', 'tolerance': {'atol': 1e-05, 'rtol': 0.0001}, 'measured': {'max_abs_error': 0.0, 'reference_max_abs': 429.46087646484375, 'relative_l2_error': 0.0, 'cosine_similarity': 1.0}}, {'path': 'grpo_dppo_custom_loss', 'check': 'baseline_gradient', 'status': 'unverified', 'tolerance': {'atol': 1e-05, 'rtol': 0.001, 'min_cosine': 0.9999}, 'measured': {'max_abs_error': 0.0, 'reference_max_abs': None, 'relative_l2_error': 0.0, 'cosine_similarity': 1.0}}, {'path': 'grpo_dppo_custom_loss', 'check': 'reference_loss', 'status': 'passed', 'tolerance': {'atol': 1e-05, 'rtol': 0.0001}, 'measured': {'max_abs_error': 3.0517578125e-05, 'reference_max_abs': 420.54302978515625, 'relative_l2_error': 7.256e-08, 'cosine_similarity': 1.0}}, {'path': 'grpo_dppo_custom_loss', 'check': 'reference_gradient', 'status': 'unverified', 'tolerance': {'atol': 1e-05, 'rtol': 0.001, 'min_cosine': 0.9999}, 'measured': {'max_abs_error': 0.0, 'reference_max_abs': None, 'relative_l2_error': 0.0, 'cosine_similarity': 1.0}}], 'reason': 'grpo_dppo_custom_loss: measured baseline_gradient errors; grpo_dppo_custom_loss: measured reference_gradient errors', 'evidence_source': 'executor report; remote execution is not attested', 'applicability': 'required', 'reported_verdict': 'inconclusive'}

decision_validation

{'status': 'uncertain', 'issues': ['Decision evidence was not recorded.'], 'criteria': [], 'outcome_reason': 'insufficient_evidence', 'unit': 'fixed_checkpoint_items', 'n_seeds': 1, 'alpha': 0.05, 'method': 'paired_hoeffding', 'applicability': {'check': 'decision_rule', 'status': 'required', 'enforce': True, 'reasons': ['experiment_features: paired_comparison is true'], 'features': {'paired_comparison': True, 'adaptive_selection': False}}}

feasibility

{'status': 'unknown', 'problems': ['Re-projection required: dataset changed or was not recorded', 'Re-projection required: steps changed or was not recorded', 'Re-projection required: batch_size changed or was not recorded', 'Re-projection required: max_seq_len changed or was not recorded', 'Re-projection required: arms changed or was not recorded', 'Re-projection required: n_examples changed or was not recorded', 'Primary endpoint changed or was not recorded', 'Unknown timing phase: sampling', 'Invalid measured timing/counts for evaluation'], 'wall_seconds': None, 'gpu_hours': None, 'known_gpu_hours': 0, 'phases': {'setup': {'wall_seconds': 153.31, 'planned_units': 1, 'planned_samples': 0.0, 'gpu_hours': None}, 'training': {'wall_seconds': 195.0, 'planned_units': 16, 'planned_samples': 512.0, 'gpu_hours': None}, 'recovery': {'wall_seconds': 249.2, 'gpu_hours': None}}, 'generation_counts': None, 'generation_applicability': {'check': 'generation_audit', 'status': 'not_applicable', 'enforce': False, 'reasons': ['spec.generation_audit opts out and no feature it covers is true'], 'features': {'generated_training_data': False}}, 'record': {'config': {'base_model': 'Qwen/Qwen3.5-2B', 'lora_rank': 32, 'arm': 'dppo_diag', 'loss': 'dppo_binary_tv', 'delta': 0.15, 'clip_ratio_c': 10.0, 'ppo_eps_counterfactual': 0.2, 'optimizer': {'type': 'adamw', 'learning_rate': 2e-05, 'beta1': 0.9, 'beta2': 0.95, 'eps': 1e-08, 'weight_decay': 0.0, 'grad_clip_norm': 0.0}, 'renderer': 'qwen3_5_disable_thinking', 'temperature': 1.0, 'max_tokens': 512, 'groups': 32, 'group_size': 8, 'data_seed': 0, 'batch_index': 0, 'minibatches': 8, 'groups_per_minibatch': 4, 'epochs': 2, 'max_updates': 16, 'loss_normalization': 'token sum (Spindle stock _loss convention)', 'init_state_path': 'tinker://21e32976f18f971463410adff4e364e4/weights/rx19badd842c_init', 'endpoint': 'shared Spindle (TINKER_BASE_URL)', 'sdk': {'tinker': '0.37.0', 'torch': '2.15.0'}, 'wall_cap_seconds': 720, 'n_eval': 0}, 'effective_config': {'base_model': 'Qwen/Qwen3.5-2B', 'lora_rank': 32, 'arm': 'dppo_diag', 'loss': 'dppo_binary_tv', 'delta': 0.15, 'clip_ratio_c': 10.0, 'ppo_eps_counterfactual': 0.2, 'optimizer': {'type': 'adamw', 'learning_rate': 2e-05, 'beta1': 0.9, 'beta2': 0.95, 'eps': 1e-08, 'weight_decay': 0.0, 'grad_clip_norm': 0.0}, 'renderer': 'qwen3_5_disable_thinking', 'temperature': 1.0, 'max_tokens': 512, 'groups': 32, 'group_size': 8, 'data_seed': 0, 'batch_index': 0, 'minibatches': 8, 'groups_per_minibatch': 4, 'epochs': 2, 'max_updates': 16, 'loss_normalization': 'token sum (Spindle stock _loss convention)', 'init_state_path': 'tinker://21e32976f18f971463410adff4e364e4/weights/rx19badd842c_init', 'endpoint': 'shared Spindle (TINKER_BASE_URL)', 'sdk': {'tinker': '0.37.0', 'torch': '2.15.0'}, 'wall_cap_seconds': 720, 'n_eval': 0}, 'primary_endpoint': 'gsm8k_test_greedy_exact_match_accuracy@final_checkpoint (1319 items) — not evaluated in this diagnostic', 'source': 'diag-events.jsonl', 'measurements': {'setup': {'seconds': 153.31, 'units': 1, 'planned_units': 1, 'samples': 0, 'gpu_count': None, 'allocation_source': None}, 'sampling': {'seconds': 114.19, 'units': 256, 'planned_units': 256, 'samples': 256, 'gpu_count': None, 'allocation_source': None}, 'training': {'seconds': 195.0, 'units': 16, 'planned_units': 16, 'samples': 512, 'gpu_count': None, 'allocation_source': None}, 'evaluation': {'seconds': 0, 'units': 0, 'planned_units': 0, 'samples': 0, 'gpu_count': None, 'allocation_source': None}}, 'generation': None, 'recovery': {'seconds': 249.2, 'gpu_count': None, 'allocation_source': None}}, 'recorded_at': 1793746408.889922}

feasibility_required

True

generation_audit_validation

{'status': 'not_applicable', 'reasons': ['spec.generation_audit opts out and no feature it covers is true'], 'applicability': {'check': 'generation_audit', 'status': 'not_applicable', 'enforce': False, 'reasons': ['spec.generation_audit opts out and no feature it covers is true'], 'features': {'generated_training_data': False}}}

initialization_provenance

{'version': 1, 'arms': [{'arm': 'dppo_diag', 'initialization': {'base_checkpoint': 'Qwen/Qwen3.5-2B', 'checkpoint_uri': 'tinker://21e32976f18f971463410adff4e364e4/weights/rx19badd842c_init', 'loaded_checkpoint_uri': 'tinker://21e32976f18f971463410adff4e364e4/weights/rx19badd842c_init', 'load_succeeded': True, 'source': 'server_response', 'checkpoint_scope': 'full_initialization', 'checkpoint_immutable': None, 'adapter_sha256': None, 'evidence_ref': 'diag-events.jsonl stage=restored (model_id daecfa24-b646-4b82-b337-4447b0e95af0:train:0)', 'before_first_update': True}, 'restore_count': 1, 'optimizer_checks': [{'after_restore': 1, 'before_next_update': True, 'requested': 'adam', 'requested_config': {'learning_rate': 2e-05, 'beta1': 0.9, 'beta2': 0.95, 'eps': 1e-08}, 'actual': 'adamw', 'actual_config': None, 'source': 'server_response', 'evidence_ref': 'diag-events.jsonl stage=restored get_info optimizer_config'}], 'initial_output': {'available': True, 'batch_sha256': '768e9246af697da5c82bfa9ea0c6a21963d5c47371d2e731c99eae9bccd21c10', 'output_sha256': '10e156f726b8feacca2c615c4257f50c7cfa06f63a56b676bd9dfbb60806c1da', 'scope': 'float32 trainer logprobs on loss tokens of 8 sentinel datums (not full logits)', 'evidence_ref': 'diag-events.jsonl stage=sentinel tag=init', 'before_first_update': True}}], 'note': 'Single diagnostic arm; no paired group. Checkpoint immutability and adapter hash not server-verifiable; optimizer hyperparameters only declared per optim_step.'}

launch

{'experiment_id': 'exp_e6b0965af3', 'context_status': 'snapshotted', 'superseded': False, 'mode': 'redo', 'purpose': "Repeat the selected saved experiment version to check reproducibility. This repetition tests that version's hypothesis and methods."}

launch_instructions

{'name': 'launch-instructions-b42946cbc224.md', 'sha256': 'b42946cbc2240a8b8fe9329beb2d97f7b8a51be4af735c400d504c87b0fead51', 'url': 'https://ta-01m6fva5db548f5skm214fepcr-8000-oah9l7x00jpo7186eb6cix5v0.w.modal.host/api/runs/run_21b1618aa7/evidence/b42946cbc2240a8b8fe9329beb2d97f7b8a51be4af735c400d504c87b0fead51'}

reported_outcome_reason

insufficient_evidence

reported_verdict

inconclusive

restoration_validation

{'status': 'unverified', 'required': True, 'checks': [{'check': 'grpo_seed0', 'status': 'unverified', 'reason': 'grpo_seed0: missing or duplicate arm evidence.'}, {'check': 'dppo_seed0', 'status': 'unverified', 'reason': 'dppo_seed0: missing or duplicate arm evidence.'}], 'reason': 'grpo_seed0: missing or duplicate arm evidence.', 'applicability': 'required'}