2026-07-11_r18_five_seed_diagnostic.json

{ "schema_version": "oczy/execution-report/v1", "record_type": "r18_five_seed_diagnostic_adjudication", "adjudication_date": "2026-07-11", "job_name": "r18-distillation-5seed-diagnostic", "module": "oczy.experiments.consolidation_distillation", "provider": "kaggle", "profile": "cpu", "source_commit": "5b5e93c63d769fea7854073a4e6c359e5d36606f", "source_repo": "https://github.com/KinoThe-Kafkaesque/oczy.git", "command": ["/usr/bin/python3", "-m", "oczy.experiments.consolidation_distillation", "--seeds", "5", "--max-steps", "10", "--stage", "stage_0_grounding"], "arguments": ["--seeds", "5", "--max-steps", "10", "--stage", "stage_0_grounding"], "status": "complete", "exit_code": 0, "error": null, "started_utc": "2026-07-11T14:03:13Z", "finished_utc": "2026-07-11T15:04:37Z", "timeout_seconds": null,

"kernel": { "kernel_id": "abdellahkadem/oczy-r18-5seed-5b5e93c63d76", "kernel_title": "Oczy R18 5seed 5b5e93c63d76", "source_dataset": "abdellahkadem/oczy-source-5b5e93c63d76", "source_archive_sha256": "bc1ff926bc679fc26e5f20cfcb0756339b002ff3c1027eb3c24251fe2f6d7f72" },

"hardware": { "platform": "Linux-6.12.90+-x86_64-with-glibc2.35", "python": "3.12.13", "torch": "2.10.0+cpu", "cuda_available": false, "cuda_device_count": 0, "torch_cuda_version": null },

"cpu_only_contract": { "cuda_available": false, "cuda_visible_devices": "", "hf_hub_offline": "1", "mkl_num_threads": "4", "oczy_remote_cpu_only": "1", "omp_num_threads": "4", "transformers_offline": "1" },

"model_artifact": { "kind": "kaggle_attached_dataset", "model_source": "qwen-lm/qwen2.5/transformers/0.5b-instruct/1", "hf_model_dir": "/kaggle/input/models/qwen-lm/qwen2.5/transformers/0.5b-instruct/1" },

"lora_config": { "lora_alpha": 16.0, "lora_lr": 0.005, "lora_rank": 8.0, "max_steps": 10, "svd_target": 0.01, "stage": "stage_0_grounding", "seeds": 5 },

"aggregate_metrics": { "distill_delta_holdout": 0.26666666666666666, "distill_specificity_delta": 0.02608695652173913, "positive_holdout_delta_count": 4, "null_holdout_delta_count": 1, "null_seeds": [2], "positive_seeds": [0, 1, 3, 4] },

"per_seed_metrics": [ { "seed": 0, "distill_delta_holdout": 0.3333333333333333, "vanilla_holdout_acc": 0.0, "lora_holdout_acc": 0.3333333333333333, "teacher_dev_delta": 0.17647058823529413, "specificity_delta": 0.0, "persistent_bytes": 17699231, "wall_s": 730.4152231329999, "holdout_delta_positive": true }, { "seed": 1, "distill_delta_holdout": 0.3333333333333333, "vanilla_holdout_acc": 0.0, "lora_holdout_acc": 0.3333333333333333, "teacher_dev_delta": 0.17647058823529413, "specificity_delta": 0.0, "persistent_bytes": 17699329, "wall_s": 728.6607881289999, "holdout_delta_positive": true }, { "seed": 2, "distill_delta_holdout": 0.0, "vanilla_holdout_acc": 0.0, "lora_holdout_acc": 0.0, "teacher_dev_delta": 0.17647058823529413, "specificity_delta": 0.043478260869565216, "persistent_bytes": 17699397, "wall_s": 728.3512674170001, "holdout_delta_positive": false }, { "seed": 3, "distill_delta_holdout": 0.3333333333333333, "vanilla_holdout_acc": 0.0, "lora_holdout_acc": 0.3333333333333333, "teacher_dev_delta": 0.17647058823529413, "specificity_delta": 0.0, "persistent_bytes": 17699383, "wall_s": 727.6332641600002, "holdout_delta_positive": true }, { "seed": 4, "distill_delta_holdout": 0.3333333333333333, "vanilla_holdout_acc": 0.0, "lora_holdout_acc": 0.3333333333333333, "teacher_dev_delta": 0.17647058823529413, "specificity_delta": 0.08695652173913043, "persistent_bytes": 17699359, "wall_s": 728.5650060029998, "holdout_delta_positive": true } ],

"teacher_validity_gate": { "gate_name": "teacher_dev_delta >= 0.2", "registered_threshold": 0.2, "threshold_changed": false, "observed_teacher_dev_delta": 0.17647058823529413, "observed_identical_across_all_seeds": true, "gate_passed": false, "margin_below_gate": 0.023529411764705884, "gate_comparison": "0.17647058823529413 < 0.2 → FAILED" },

"scientific_classification": { "classification": "BLOCKED", "classification_detail": "BLOCKED at teacher validity gate / diagnostic only", "h_distill_verdict_permitted": false, "h_distill_verdict_reason": "Teacher gate failed after registered fallback. The prior 3-seed stage_0 run produced teacher_dev_delta=0.1765 (below 0.2 gate). The 5-seed rerun was registered as a diagnostic fallback and also produced teacher_dev_delta=0.17647058823529413 across all 5 seeds. No H-DISTILL verdict is permitted because the teacher validity gate did not pass.", "positive_signal_visible": true, "positive_signal_detail": "4 of 5 seeds show positive distill_delta_holdout (0.3333333333333333). This is conditional signal only — it cannot be elevated to a scientific verdict while the teacher gate is failed.", "null_signal_visible": true, "null_signal_detail": "Seed 2 shows distill_delta_holdout=0.0 (null). The seed-2 null is visible alongside the 4/5 positive signal.", "infrastructure_exit_code": 0, "infrastructure_succeeded": true, "infrastructure_vs_scientific_separation": "Infrastructure completed successfully (exit 0, all 5 seeds ran to completion). This is distinct from the scientific classification, which is BLOCKED at the teacher validity gate. Infrastructure success does not imply a positive scientific verdict." },

"provenance": { "launch_record": "experiments_logs/2026-07-11_live_runner_queue.json", "local_evidence_log": "/tmp/oczy-live-queue/reports/r18-distillation-5seed-diagnostic/oczy-r18-5seed-5b5e93c63d76.log", "local_provenance_json": "/tmp/oczy-live-queue/reports/r18-distillation-5seed-diagnostic/remote_run_provenance.json", "local_state_json": "/tmp/oczy-live-queue/state.json", "scheduler_state": "succeeded", "scheduler_completed_at_epoch": 1783782292.9804573, "scheduler_completed_at_utc": "2026-07-11T15:04:52Z", "scheduler_collected_at_epoch": 1783782294.7386103, "scheduler_collected_at_utc": "2026-07-11T15:04:54Z", "scheduler_submitted_at_epoch": 1783778582.3723054, "scheduler_submitted_at_utc": "2026-07-11T14:03:02Z", "source_manifest": { "archive_sha256": "bc1ff926bc679fc26e5f20cfcb0756339b002ff3c1027eb3c24251fe2f6d7f72", "archive_size_bytes": 1685358, "tracked_file_count": 453, "branch_at_packaging": "autoresearch/session-20260625", "created_utc": "2026-07-11T14:02:03Z", "dataset_id": "abdellahkadem/oczy-source-5b5e93c63d76", "dirty_entry_count": 0, "worktree_dirty_at_packaging": false } },

"historical_context": { "prior_3seed_run": "The prior 3-seed stage_0 run (Campaign 0d48130, commit 537260c) produced teacher_dev_delta=0.1765 (below the 0.2 gate). It was infrastructure-complete (2/3 seeds showed positive distill_delta_holdout, 1/3 null) but scientifically BLOCKED because teacher_dev_delta=0.1765 < 0.2; the prior PARTIAL label was corrected during adjudication. See experiments_logs/2026-07-11_campaign_0d48130.md.", "registered_fallback": "The 5-seed stage_0 rerun was registered as a diagnostic fallback to check whether the teacher gate would pass with more seeds. It did not: teacher_dev_delta=0.17647058823529413 is identical across all 5 seeds and remains below the 0.2 gate.", "prior_r18_gate_result": "The single-seed R18 gate run (Campaign 0d48130) was infrastructure-complete with distill_delta_holdout=0.3333 but scientifically BLOCKED because teacher_dev_delta=0.1765 < 0.2; the prior gate-passed label was corrected during adjudication. The full experiment requires the teacher validity gate to pass.", "campaign_log": "experiments_logs/2026-07-11_campaign_0d48130.md" },

"explicit_non_claims": [ "No positive scientific verdict is claimed. The classification is BLOCKED at the teacher validity gate.", "The 4/5 positive holdout deltas are conditional signal only and cannot be elevated to an H-DISTILL verdict while the teacher gate is failed.", "The seed-2 null (distill_delta_holdout=0.0) is reported as observed; it is not interpreted as a refutation.", "No threshold, metric, baseline, episode, scoring, eval manifest, or research spec was changed.", "The teacher_dev_delta >= 0.2 gate is unchanged and was not moved to accommodate the observed 0.1765.", "Infrastructure success (exit 0) is distinct from scientific success (gate passed)." ],

"scope_limit": "Five-seed stage_0_grounding diagnostic run on Kaggle CPU (Qwen2.5-0.5B-Instruct, LoRA rank 8, alpha 16, lr 0.005, max_steps 10). No threshold, metric, baseline, episode, scoring, eval manifest, or research spec was changed. The teacher_dev_delta >= 0.2 validity gate is unchanged." }