{ "audit_scope": "Read-only Meta reconciliation of current bounded backfill and identified ledger pending records; no code edits/training/download/PR; official paper HTML used only for three targeted resource checks.", "repository": ".", "summary": { "backfill_new_total": 54, "backfill_new_meta": 0, "backfill_new_google": 54, "meta_identities_in_backfill": 19, "meta_backfill_status_counts": { "reviewed": 15, "implemented": 4 }, "all_current_computed_implementation_pending": 521, "identified_meta_related_pending": 3, "meta_directory_total": 48, "meta_directory_decision_counts": { "preserve_existing_catalog": 6, "preserve_prior_scope_exclusion": 16, "preserve_existing_diagnostic_and_repair_source_date_linkage; full original experiment not reproduced high cost": 1, "archive_metadata_only_unconfirmed_nonblocking": 21, "recommend_reference_catalog_not_reproduced_high_cost": 3, "recommend_reference_catalog_not_reproduced_high_cost; affordable task-metadata/result-import interface possible, not currently integrated": 1 }, "known_deferred_meta_cost_closed": 9 }, "critical_interpretations": [ "54 is a bounded source-ID status count, not all Google+Meta publications and not 54 approved implementations. All54 currently come from Google signals.", "The48 Meta official-directory rows include21 unconfirmed metadata-only URL identities and16 scope exclusions. Null-ID rows are absent from the83-ID backfill count; do not call them fully reviewed or implementation TODOs.", "The4 Meta implemented statuses are L1 diagnostics, with diagnostic_only true and formal_comparison false. PAHF aggregate seed_mean43 is not capability evidence.", "The15 Meta reviewed IDs are terminal high-cost references or preserved diagnostic+cost closures. Do not revive them as reproduction TODOs.", "The521 number is reproducible as repository_paper_statuses output, not established as521 real unimplemented affordable papers. Three identified Meta-related rows remain pending due old acceptance records; targeted primary checks expose missing cost disposition, not a new mandate to implement.", "This is bounded reconciliation, not exhaustive identification of every Meta-authored paper among521; affiliation data are incomplete and broad query hits are not authorship proof." ], "status_algorithm_evidence": [ "scripts/audit_historical_priority_backfill.py:23", "scripts/audit_historical_priority_backfill.py:63", "src/auto_research/discovery.py:153", "src/auto_research/discovery.py:159", "src/auto_research/discovery_review.py:65", "src/auto_research/discovery_review.py:83" ], "pending_rows": [ { "id": "2610.06064", "title": "TrustMI: Causally controlling how assistants trust their users", "current_computed_status": "implementation-pending", "judgment": "pending_implementation_exists_but_cost_gate_missing_not_authorized_implementation_queue", "official_source": "https://arxiv.org/html/2610.06064v1", "primary_resource_evidence": "Official original §4, Appendix C/D.2/D.4: six 7B-70B models; learned steering matrices trained 10 epochs, 120B judge on H100, multiple agent and capability benchmarks, mostly three repetitions. Full campaign should not be treated as an affordable automatic TODO; smaller faithful scope requires explicit cost assessment.", "primary_locations": "§4; Appendix C; Appendix D.2-D.4 (web lines 164-170, 763-802)", "evidence_paths": [ "docs/paper-discovery-ledger.json:22928", "docs/paper-discovery-ledger.json:22943", "docs/research-audits/oct07-review-closure/paper-register.json:13351", "docs/research-audits/oct07-review-closure/paper-register.json:13357", "docs/research-audits/oct07-review-closure/paper-register.json:13362", "docs/research-audits/oct12-review-closure/paper-register.json:41893", "docs/research-audits/oct12-review-closure/paper-register.json:41899", "docs/research-audits/oct12-review-closure/paper-register.json:41904", "docs/research-audits/oct15-review-closure/paper-register.json:70336" ], "manifest_match": null, "implementation_search_result": "Exact ID/title/name search in docs/src/tests found review artifacts only, no reproduction directory, adapter, runnable paper implementation or local metrics for this identity.", "fulltext_review_not_implementation": "Oct15 inherits known status, explicitly says existing_evidence_not_upgraded_by_review; this does not close previous accepted/not_implemented delivery." }, { "id": "2610.04721", "title": "Knossos and Ariadne: Benchmarking and Learning Complete Diagram Topology Extraction with Vision-Language Models", "current_computed_status": "implementation-pending", "judgment": "stale_pending_original_resource_evidence_supports_high_cost_terminal_not_auto_TODO", "official_source": "https://arxiv.org/html/2610.04721v1", "primary_resource_evidence": "Official original Appendix E.1/Table10: NODE1125, EDGE5960, TRACE5801 steps for each 8B/9B backbone; 4 A100 GPUs, QLoRA. Full original reproduction exceeds single A100/A30 campaign budget; ledger has not recorded terminal cost closure.", "primary_locations": "Appendix E.1 Table10 (web lines 681-693)", "evidence_paths": [ "docs/paper-discovery-ledger.json:27797", "docs/paper-discovery-ledger.json:27811", "docs/research-audits/oct07-review-closure/paper-register.json:19671", "docs/research-audits/oct07-review-closure/paper-register.json:19677", "docs/research-audits/oct07-review-closure/paper-register.json:19689", "docs/research-audits/oct07-review-closure/paper-register.json:19693", "docs/research-audits/oct07-review-closure/paper-register.json:19697", "docs/research-audits/oct07-review-closure/paper-register.json:19703", "docs/research-audits/oct12-review-closure/paper-register.json:47902", "docs/research-audits/oct12-review-closure/paper-register.json:47908", "docs/research-audits/oct12-review-closure/paper-register.json:47920", "docs/research-audits/oct12-review-closure/paper-register.json:47924", "docs/research-audits/oct12-review-closure/paper-register.json:47928", "docs/research-audits/oct12-review-closure/paper-register.json:47934", "docs/research-audits/oct15-review-closure/paper-register.json:75420" ], "manifest_match": null, "implementation_search_result": "Exact ID/title/name search in docs/src/tests found review artifacts only, no reproduction directory, adapter, runnable paper implementation or local metrics for this identity.", "fulltext_review_not_implementation": "Oct15 inherits known status, explicitly says existing_evidence_not_upgraded_by_review; this does not close previous accepted/not_implemented delivery." }, { "id": "2610.09484", "title": "MIMESIS: Learning User Simulators as Training Environments for Interactive Agents", "current_computed_status": "implementation-pending", "judgment": "stale_pending_original_resource_evidence_supports_high_cost_terminal_not_auto_TODO", "official_source": "https://arxiv.org/html/2610.09484v1", "primary_resource_evidence": "Official original A.1/Table4: mid-training32 A100 for about4.8days, 9.51B tokens; SFT8 A100; A.3/Table7 simulator RL16 A100/500steps/512rollouts; D.1/Table13 actor RL16 A100. No repository cost terminal is recorded.", "primary_locations": "Appendix A.1 Table4, A.3 Table7, D.1 Table13 (web lines 304-328, 382-427, 658-678)", "evidence_paths": [ "docs/paper-discovery-ledger.json:45040", "docs/research-audits/oct12-review-closure/paper-register.json:27428", "docs/research-audits/oct12-review-closure/paper-register.json:27433", "docs/research-audits/oct12-review-closure/paper-register.json:27463", "docs/research-audits/oct12-review-closure/paper-register.json:27469", "docs/research-audits/oct15-review-closure/paper-register.json:116978" ], "manifest_match": null, "implementation_search_result": "Exact ID/title/name search in docs/src/tests found review artifacts only, no reproduction directory, adapter, runnable paper implementation or local metrics for this identity.", "fulltext_review_not_implementation": "Oct15 inherits known status, explicitly says existing_evidence_not_upgraded_by_review; this does not close previous accepted/not_implemented delivery." } ], "meta_audit_rows": [ { "id": "2610.24628", "title": "Muse Glimmer 30B: An On-Device Agentic Model", "source": "https://ai.meta.com/research/publications/muse-glimmer-30b-an-on-device-agentic-model/", "recorded_decision": "preserve_existing_catalog", "current_computed_status": "reviewed", "judgment": "terminal_high_cost_review_DONE_not_reproduced_no_automatic_retry", "full_article_read": false, "review_level": "catalog_linkage_check", "cost_basis": null, "evidence_paths": [ "docs/research-audits/harness2-priority-repair/meta-audit.json:23", "docs/paper-discovery-ledger.json:122215", "docs/paper-discovery-ledger.json:122234", "docs/paper-discovery-ledger.json:122238", "docs/paper-discovery-ledger.json:122252", "docs/research-audits/harness2-priority-repair/backfill-inventory.json:787", "docs/research-manifest.json:8536", "docs/research-manifest.json:8537", "docs/foundation-models/2610.24628-muse-glimmer-30b/README.md:3", "docs/foundation-models/2610.24628-muse-glimmer-30b/README.md:60" ], "recorded_primary_locations": [ "not in Oct15 Meta frozen source review", "/tmp/oct20-source-delta/retrieval.json: Meta page 1, checked 2026-10-20T06:08:16.356765+00:00" ], "catalog_key": "muse-glimmer-30b", "detail_path": "docs/foundation-models/2610.24628-muse-glimmer-30b/README.md", "catalog_implementation_status": "deferred_high_cost" }, { "id": null, "title": "Human-Guided Agentic AI for End-to-End EHR Research: Sequential Antibody-Drug Conjugate Pathways in Breast Cancer", "source": "https://ai.meta.com/research/publications/human-guided-agentic-ai-for-end-to-end-ehr-research-sequential-antibody-drug-conjugate-pathways-in-breast-cancer/", "recorded_decision": "preserve_prior_scope_exclusion", "current_computed_status": null, "judgment": "scope_excluded_not_reproduction_TODO", "full_article_read": false, "review_level": "inherited_scope_evidence_not_new_fulltext_review", "cost_basis": null, "evidence_paths": [ "docs/research-audits/harness2-priority-repair/meta-audit.json:47" ], "recorded_primary_locations": [ "docs/research-audits/oct15-review-closure/meta-source-review.json: exact URL record", "/tmp/oct20-source-delta/retrieval.json: Meta page 1, checked 2026-10-20T06:08:16.356765+00:00" ] }, { "id": null, "title": "Pickleball injuries, search interest, and participation in the United States, 2014-2023: an ecological time series study", "source": "https://ai.meta.com/research/publications/pickleball-injuries-search-interest-and-participation-in-the-united-states-2014-2023-an-ecological-time-series-study/", "recorded_decision": "preserve_prior_scope_exclusion", "current_computed_status": null, "judgment": "scope_excluded_not_reproduction_TODO", "full_article_read": false, "review_level": "inherited_scope_evidence_not_new_fulltext_review", "cost_basis": null, "evidence_paths": [ "docs/research-audits/harness2-priority-repair/meta-audit.json:70" ], "recorded_primary_locations": [ "docs/research-audits/oct15-review-closure/meta-source-review.json: exact URL record", "/tmp/oct20-source-delta/retrieval.json: Meta page 1, checked 2026-10-20T06:08:16.356765+00:00" ] }, { "id": null, "title": "Outcomes in Methicillin-Susceptible and Methicillin-Resistant Staphylococcus aureus Infective Endocarditis Compared With Staphylococcus lugdunensis Infective Endocarditis", "source": "https://ai.meta.com/research/publications/outcomes-in-methicillin-susceptible-and-methicillin-resistant-staphylococcus-aureus-infective-endocarditis-compared-with-staphylococcus-lugdunensis-infective-endocarditis/", "recorded_decision": "preserve_prior_scope_exclusion", "current_computed_status": null, "judgment": "scope_excluded_not_reproduction_TODO", "full_article_read": false, "review_level": "inherited_scope_evidence_not_new_fulltext_review", "cost_basis": null, "evidence_paths": [ "docs/research-audits/harness2-priority-repair/meta-audit.json:93" ], "recorded_primary_locations": [ "docs/research-audits/oct15-review-closure/meta-source-review.json: exact URL record", "/tmp/oct20-source-delta/retrieval.json: Meta page 1, checked 2026-10-20T06:08:16.356765+00:00" ] }, { "id": null, "title": "Social Needs Z Codes and Subsequent Acute Care Across 21 U.S. Health Systems", "source": "https://ai.meta.com/research/publications/social-needs-z-codes-and-subsequent-acute-care-across-21-u-s-health-systems/", "recorded_decision": "preserve_prior_scope_exclusion", "current_computed_status": null, "judgment": "scope_excluded_not_reproduction_TODO", "full_article_read": false, "review_level": "inherited_scope_evidence_not_new_fulltext_review", "cost_basis": null, "evidence_paths": [ "docs/research-audits/harness2-priority-repair/meta-audit.json:116" ], "recorded_primary_locations": [ "docs/research-audits/oct15-review-closure/meta-source-review.json: exact URL record", "/tmp/oct20-source-delta/retrieval.json: Meta page 1, checked 2026-10-20T06:08:16.356765+00:00" ] }, { "id": null, "title": "Cladophialophora bantiana: a US clinical case series assisted by artificial intelligence", "source": "https://ai.meta.com/research/publications/cladophialophora-bantiana-a-us-clinical-case-series-assisted-by-artificial-intelligence/", "recorded_decision": "preserve_prior_scope_exclusion", "current_computed_status": null, "judgment": "scope_excluded_not_reproduction_TODO", "full_article_read": false, "review_level": "inherited_scope_evidence_not_new_fulltext_review", "cost_basis": null, "evidence_paths": [ "docs/research-audits/harness2-priority-repair/meta-audit.json:139" ], "recorded_primary_locations": [ "docs/research-audits/oct15-review-closure/meta-source-review.json: exact URL record", "/tmp/oct20-source-delta/retrieval.json: Meta page 1, checked 2026-10-20T06:08:16.356765+00:00" ] }, { "id": null, "title": "Trends in Emergency Department Visits for Electric Scooter and Electric Bicycle Injuries in the United States, 2014–2023: A Retrospective Observational Study", "source": "https://ai.meta.com/research/publications/trends-in-emergency-department-visits-for-electric-scooter-and-electric-bicycle-injuries-in-the-united-states-2014-2023-a-retrospective-observational-study/", "recorded_decision": "preserve_prior_scope_exclusion", "current_computed_status": null, "judgment": "scope_excluded_not_reproduction_TODO", "full_article_read": false, "review_level": "inherited_scope_evidence_not_new_fulltext_review", "cost_basis": null, "evidence_paths": [ "docs/research-audits/harness2-priority-repair/meta-audit.json:162" ], "recorded_primary_locations": [ "docs/research-audits/oct15-review-closure/meta-source-review.json: exact URL record", "/tmp/oct20-source-delta/retrieval.json: Meta page 1, checked 2026-10-20T06:08:16.356765+00:00" ] }, { "id": null, "title": "Documented Dental Disease in Patients With Staphylococcus aureus Bacteremia", "source": "https://ai.meta.com/research/publications/documented-dental-disease-in-patients-with-staphylococcus-aureus-bacteremia/", "recorded_decision": "preserve_prior_scope_exclusion", "current_computed_status": null, "judgment": "scope_excluded_not_reproduction_TODO", "full_article_read": false, "review_level": "inherited_scope_evidence_not_new_fulltext_review", "cost_basis": null, "evidence_paths": [ "docs/research-audits/harness2-priority-repair/meta-audit.json:185" ], "recorded_primary_locations": [ "docs/research-audits/oct15-review-closure/meta-source-review.json: exact URL record", "/tmp/oct20-source-delta/retrieval.json: Meta page 1, checked 2026-10-20T06:08:16.356765+00:00" ] }, { "id": null, "title": "Temporal Patterns of Arrhythmia and Conduction Disorder Diagnoses After Infective Endocarditis", "source": "https://ai.meta.com/research/publications/temporal-patterns-of-arrhythmia-and-conduction-disorder-diagnoses-after-infective-endocarditis/", "recorded_decision": "preserve_prior_scope_exclusion", "current_computed_status": null, "judgment": "scope_excluded_not_reproduction_TODO", "full_article_read": false, "review_level": "inherited_scope_evidence_not_new_fulltext_review", "cost_basis": null, "evidence_paths": [ "docs/research-audits/harness2-priority-repair/meta-audit.json:208" ], "recorded_primary_locations": [ "docs/research-audits/oct15-review-closure/meta-source-review.json: exact URL record", "/tmp/oct20-source-delta/retrieval.json: Meta page 1, checked 2026-10-20T06:08:16.356765+00:00" ] }, { "id": null, "title": "Remission of Type 2 Diabetes in Linked Health Records: Uncommon, Short-Lived, and More Often Ended by Treatment Restart Than by Rising HbA1c", "source": "https://ai.meta.com/research/publications/remission-of-type-2-diabetes-in-linked-health-records-uncommon-short-lived-and-more-often-ended-by-treatment-restart-than-by-rising-hba1c/", "recorded_decision": "preserve_prior_scope_exclusion", "current_computed_status": null, "judgment": "scope_excluded_not_reproduction_TODO", "full_article_read": false, "review_level": "inherited_scope_evidence_not_new_fulltext_review", "cost_basis": null, "evidence_paths": [ "docs/research-audits/harness2-priority-repair/meta-audit.json:231" ], "recorded_primary_locations": [ "docs/research-audits/oct15-review-closure/meta-source-review.json: exact URL record", "/tmp/oct20-source-delta/retrieval.json: Meta page 1, checked 2026-10-20T06:08:16.356765+00:00" ] }, { "id": null, "title": "Return Visits, Admission, and Mortality After Leaving Without Being Seen Among Emergency Department Patients With Behavioral Health Presentations", "source": "https://ai.meta.com/research/publications/return-visits-admission-and-mortality-after-leaving-without-being-seen-among-emergency-department-patients-with-behavioral-health-presentations/", "recorded_decision": "preserve_prior_scope_exclusion", "current_computed_status": null, "judgment": "scope_excluded_not_reproduction_TODO", "full_article_read": false, "review_level": "inherited_scope_evidence_not_new_fulltext_review", "cost_basis": null, "evidence_paths": [ "docs/research-audits/harness2-priority-repair/meta-audit.json:254" ], "recorded_primary_locations": [ "docs/research-audits/oct15-review-closure/meta-source-review.json: exact URL record", "/tmp/oct20-source-delta/retrieval.json: Meta page 1, checked 2026-10-20T06:08:16.356765+00:00" ] }, { "id": null, "title": "AutoformBench: A Lean Formalization Benchmark for Advanced Mathematics", "source": "https://ai.meta.com/research/publications/autoformbench-a-lean-formalization-benchmark-for-advanced-mathematics/", "recorded_decision": "preserve_existing_catalog", "current_computed_status": null, "judgment": "existing_reference_catalog_high_cost_not_reproduced_identity_not_2603.29832", "full_article_read": false, "review_level": "catalog_linkage_check", "cost_basis": null, "evidence_paths": [ "docs/research-audits/harness2-priority-repair/meta-audit.json:277", "docs/agent-research/autoformbench/README.md:148" ], "recorded_primary_locations": [ "docs/research-audits/oct15-review-closure/meta-source-review.json: exact URL record", "/tmp/oct20-source-delta/retrieval.json: Meta page 1, checked 2026-10-20T06:08:16.356765+00:00" ], "catalog_key": "autoformbench", "detail_path": "docs/agent-research/autoformbench/README.md", "catalog_implementation_status": "evaluation_protocol_interface" }, { "id": null, "title": "Tightness of the Cycle-Based Relaxation for Completed Length-Three Alpha-Cycles", "source": "https://ai.meta.com/research/publications/tightness-of-the-cycle-based-relaxation-for-completed-length-three-alpha-cycles/", "recorded_decision": "preserve_prior_scope_exclusion", "current_computed_status": null, "judgment": "scope_excluded_not_reproduction_TODO", "full_article_read": false, "review_level": "inherited_scope_evidence_not_new_fulltext_review", "cost_basis": null, "evidence_paths": [ "docs/research-audits/harness2-priority-repair/meta-audit.json:300" ], "recorded_primary_locations": [ "docs/research-audits/oct15-review-closure/meta-source-review.json: exact URL record", "/tmp/oct20-source-delta/retrieval.json: Meta page 2, checked 2026-10-20T06:08:54.282091+00:00" ] }, { "id": null, "title": "On Solvable Evolution Algebras and a Conjecture by García-Martínez and Pérez-Rodríguez", "source": "https://ai.meta.com/research/publications/on-solvable-evolution-algebras-and-a-conjecture-by-garcia-martinez-and-perez-rodriguez/", "recorded_decision": "preserve_prior_scope_exclusion", "current_computed_status": null, "judgment": "scope_excluded_not_reproduction_TODO", "full_article_read": false, "review_level": "inherited_scope_evidence_not_new_fulltext_review", "cost_basis": null, "evidence_paths": [ "docs/research-audits/harness2-priority-repair/meta-audit.json:323" ], "recorded_primary_locations": [ "docs/research-audits/oct15-review-closure/meta-source-review.json: exact URL record", "/tmp/oct20-source-delta/retrieval.json: Meta page 2, checked 2026-10-20T06:08:54.282091+00:00" ] }, { "id": null, "title": "String Two-Point Function = Height Function on a Curve", "source": "https://ai.meta.com/research/publications/string-two-point-function-height-function-on-a-curve/", "recorded_decision": "preserve_prior_scope_exclusion", "current_computed_status": null, "judgment": "scope_excluded_not_reproduction_TODO", "full_article_read": false, "review_level": "inherited_scope_evidence_not_new_fulltext_review", "cost_basis": null, "evidence_paths": [ "docs/research-audits/harness2-priority-repair/meta-audit.json:346" ], "recorded_primary_locations": [ "docs/research-audits/oct15-review-closure/meta-source-review.json: exact URL record", "/tmp/oct20-source-delta/retrieval.json: Meta page 2, checked 2026-10-20T06:08:54.282091+00:00" ] }, { "id": null, "title": "Semiabelian Groups Need Not Be Monomial", "source": "https://ai.meta.com/research/publications/semiabelian-groups-need-not-be-monomial/", "recorded_decision": "preserve_prior_scope_exclusion", "current_computed_status": null, "judgment": "scope_excluded_not_reproduction_TODO", "full_article_read": false, "review_level": "inherited_scope_evidence_not_new_fulltext_review", "cost_basis": null, "evidence_paths": [ "docs/research-audits/harness2-priority-repair/meta-audit.json:369" ], "recorded_primary_locations": [ "docs/research-audits/oct15-review-closure/meta-source-review.json: exact URL record", "/tmp/oct20-source-delta/retrieval.json: Meta page 2, checked 2026-10-20T06:08:54.282091+00:00" ] }, { "id": null, "title": "Finite-Time Blow-Up of Radial Negative-Energy Solutions for the Mass-Critical Biharmonic Nonlinear Schrödinger Equation", "source": "https://ai.meta.com/research/publications/finite-time-blow-up-of-radial-negative-energy-solutions-for-the-mass-critical-biharmonic-nonlinear-schrodinger-equation/", "recorded_decision": "preserve_prior_scope_exclusion", "current_computed_status": null, "judgment": "scope_excluded_not_reproduction_TODO", "full_article_read": false, "review_level": "inherited_scope_evidence_not_new_fulltext_review", "cost_basis": null, "evidence_paths": [ "docs/research-audits/harness2-priority-repair/meta-audit.json:392" ], "recorded_primary_locations": [ "docs/research-audits/oct15-review-closure/meta-source-review.json: exact URL record", "/tmp/oct20-source-delta/retrieval.json: Meta page 2, checked 2026-10-20T06:08:54.282091+00:00" ] }, { "id": null, "title": "The Strict Threshold for Gaussian Ellipsoid Fitting", "source": "https://ai.meta.com/research/publications/the-strict-threshold-for-gaussian-ellipsoid-fitting/", "recorded_decision": "preserve_prior_scope_exclusion", "current_computed_status": null, "judgment": "scope_excluded_not_reproduction_TODO", "full_article_read": false, "review_level": "inherited_scope_evidence_not_new_fulltext_review", "cost_basis": null, "evidence_paths": [ "docs/research-audits/harness2-priority-repair/meta-audit.json:415" ], "recorded_primary_locations": [ "docs/research-audits/oct15-review-closure/meta-source-review.json: exact URL record", "/tmp/oct20-source-delta/retrieval.json: Meta page 2, checked 2026-10-20T06:08:54.282091+00:00" ] }, { "id": "2609.31644", "title": "MaD-RL: Matching Distributions for Calibrating LLMs with Reinforcement Learning", "source": "https://ai.meta.com/research/publications/mad-rl-matching-distributions-for-calibrating-llms-with-reinforcement-learning/", "recorded_decision": "preserve_existing_diagnostic_and_repair_source_date_linkage; full original experiment not reproduced high cost", "current_computed_status": "reviewed", "judgment": "terminal_high_cost_review_DONE_not_reproduced_no_automatic_retry", "full_article_read": false, "review_level": "bounded_original_method_and_cost_check_plus_existing_evidence_linkage", "cost_basis": "Qwen3-4B: synthetic 6000 updates, G32, 48 runs; math/code 1200 updates, G16 across three datasets and three seeds. No new run proposed.", "evidence_paths": [ "docs/research-audits/harness2-priority-repair/meta-audit.json:437", "docs/paper-discovery-ledger.json:122730", "docs/paper-discovery-ledger.json:122746", "docs/paper-discovery-ledger.json:122747", "docs/paper-discovery-ledger.json:122757", "docs/paper-discovery-ledger.json:122764", "docs/research-audits/harness2-priority-repair/backfill-inventory.json:646", "docs/research-manifest.json:13598", "docs/research-manifest.json:13599", "docs/post-training/2609.31644-mad-rl/README.md:3", "docs/post-training/2609.31644-mad-rl/README.md:54", "docs/experiments/mad-rl-mechanism.md:78", "src/auto_research/post_training/mad_rl.py:38", "src/auto_research/post_training/mad_rl.py:47", "src/auto_research/post_training/mad_rl.py:84", "src/auto_research/post_training/mad_rl.py:90", "src/auto_research/post_training/mad_rl.py:96", "src/auto_research/post_training/mad_rl.py:103" ], "recorded_primary_locations": [ "https://arxiv.org/abs/2609.31644: Submission history v1", "https://arxiv.org/html/2609.31644v1: author affiliations, sections 2.1-2.5, 3.2 and Appendix C", "docs/experiments/mad-rl-mechanism.md: existing diagnostic and incorrect first-publication field", "src/auto_research/post_training/mad_rl.py", "docs/gpu-validations/mad-rl-qwen-choice-a100-20260929.json", "docs/experiments/evidence/mad-rl-qwen-choice-seeds42-44.json" ], "catalog_key": "mad-rl", "detail_path": "docs/post-training/2609.31644-mad-rl/README.md", "catalog_implementation_status": "deferred_high_cost", "local_evidence": "Already-linked independent GRU and A100 Qwen mechanism diagnostics; negative three-arm results retained; no formal paper adapter." }, { "id": null, "title": "Repeat-After-Me: Black-Box Adaptive Visual Prompt Injection", "source": "https://ai.meta.com/research/publications/repeat-after-me-black-box-adaptive-visual-prompt-injection/", "recorded_decision": "archive_metadata_only_unconfirmed_nonblocking", "current_computed_status": null, "judgment": "metadata_only_unresolved_identity_or_method_archived_nonblocking_not_implementation_TODO", "full_article_read": false, "review_level": "official_directory_and_frozen_history_identity_check_only", "cost_basis": null, "evidence_paths": [ "docs/research-audits/harness2-priority-repair/meta-audit.json:466" ], "recorded_primary_locations": [ "docs/research-audits/oct15-review-closure/meta-source-review.json: exact URL record", "/tmp/oct20-source-delta/retrieval.json: Meta page 2, checked 2026-10-20T06:08:54.282091+00:00" ] }, { "id": null, "title": "Alignment-Free Text-Audiobox for Voice Dubbing and Full-Duplex Dialogue Synthesis", "source": "https://ai.meta.com/research/publications/alignment-free-text-audiobox-for-voice-dubbing-and-full-duplex-dialogue-synthesis/", "recorded_decision": "archive_metadata_only_unconfirmed_nonblocking", "current_computed_status": null, "judgment": "metadata_only_unresolved_identity_or_method_archived_nonblocking_not_implementation_TODO", "full_article_read": false, "review_level": "official_directory_and_frozen_history_identity_check_only", "cost_basis": null, "evidence_paths": [ "docs/research-audits/harness2-priority-repair/meta-audit.json:489" ], "recorded_primary_locations": [ "docs/research-audits/oct15-review-closure/meta-source-review.json: exact URL record", "/tmp/oct20-source-delta/retrieval.json: Meta page 2, checked 2026-10-20T06:08:54.282091+00:00" ] }, { "id": null, "title": "WaiT for the Signal: Simple Frequency-Aware Flow-Matching", "source": "https://ai.meta.com/research/publications/wait-for-the-signal-simple-frequency-aware-flow-matching/", "recorded_decision": "archive_metadata_only_unconfirmed_nonblocking", "current_computed_status": null, "judgment": "metadata_only_unresolved_identity_or_method_archived_nonblocking_not_implementation_TODO", "full_article_read": false, "review_level": "official_directory_and_frozen_history_identity_check_only", "cost_basis": null, "evidence_paths": [ "docs/research-audits/harness2-priority-repair/meta-audit.json:512" ], "recorded_primary_locations": [ "docs/research-audits/oct15-review-closure/meta-source-review.json: exact URL record", "/tmp/oct20-source-delta/retrieval.json: Meta page 2, checked 2026-10-20T06:08:54.282091+00:00" ] }, { "id": "2607.25970", "title": "Reinforcement Learning for Code Optimization", "source": "https://ai.meta.com/research/publications/reinforcement-learning-for-code-optimization/", "recorded_decision": "recommend_reference_catalog_not_reproduced_high_cost", "current_computed_status": "reviewed", "judgment": "terminal_high_cost_review_DONE_not_reproduced_no_automatic_retry", "full_article_read": false, "review_level": "bounded_original_method_evaluation_and_resource_review", "cost_basis": "Original section 4: each online RL run uses 8-32 GPU nodes for hours or days; Qwen2.5 7B/32B and CWM32B plus calibrated remote code execution. This is far above one A100/A30.", "evidence_paths": [ "docs/research-audits/harness2-priority-repair/meta-audit.json:534", "docs/paper-discovery-ledger.json:122686", "docs/paper-discovery-ledger.json:122702", "docs/paper-discovery-ledger.json:122703", "docs/paper-discovery-ledger.json:122704", "docs/paper-discovery-ledger.json:122710", "docs/paper-discovery-ledger.json:122717", "docs/research-audits/harness2-priority-repair/backfill-inventory.json:574", "docs/research-manifest.json:12337", "docs/research-manifest.json:12338", "docs/post-training/2607.25970-code-optimization-rl/README.md:3", "docs/post-training/2607.25970-code-optimization-rl/README.md:45" ], "recorded_primary_locations": [ "docs/paper-audits/2026-historical-candidates.json: 2607.25970 P2/query-collision, full_text_review_required=false", "https://arxiv.org/abs/2607.25970: v1/v2 history", "https://arxiv.org/html/2607.25970v1: author affiliations; sections 3-5; Appendix C timing calibration, D reward composition, E offline simulator, F async GRPO" ], "catalog_key": "code-optimization-rl", "detail_path": "docs/post-training/2607.25970-code-optimization-rl/README.md", "catalog_implementation_status": "deferred_high_cost" }, { "id": null, "title": "Learning to Reason by Analogy via Retrieval-Augmented Reinforcement Fine-Tuning", "source": "https://ai.meta.com/research/publications/learning-to-reason-by-analogy-via-retrieval-augmented-reinforcement-fine-tuning/", "recorded_decision": "archive_metadata_only_unconfirmed_nonblocking", "current_computed_status": null, "judgment": "metadata_only_unresolved_identity_or_method_archived_nonblocking_not_implementation_TODO", "full_article_read": false, "review_level": "official_directory_and_frozen_history_identity_check_only", "cost_basis": null, "evidence_paths": [ "docs/research-audits/harness2-priority-repair/meta-audit.json:560" ], "recorded_primary_locations": [ "docs/research-audits/oct15-review-closure/meta-source-review.json: exact URL record", "/tmp/oct20-source-delta/retrieval.json: Meta page 2, checked 2026-10-20T06:08:54.282091+00:00" ] }, { "id": null, "title": "S-EMBER: A Large-Scale Benchmark for Streaming Egocentric Memory Retrieval", "source": "https://ai.meta.com/research/publications/s-ember-a-large-scale-benchmark-for-streaming-egocentric-memory-retrieval/", "recorded_decision": "archive_metadata_only_unconfirmed_nonblocking", "current_computed_status": null, "judgment": "metadata_only_unresolved_identity_or_method_archived_nonblocking_not_implementation_TODO", "full_article_read": false, "review_level": "official_directory_and_frozen_history_identity_check_only", "cost_basis": null, "evidence_paths": [ "docs/research-audits/harness2-priority-repair/meta-audit.json:583" ], "recorded_primary_locations": [ "docs/research-audits/oct15-review-closure/meta-source-review.json: exact URL record", "/tmp/oct20-source-delta/retrieval.json: Meta page 3, checked 2026-10-20T06:08:54.543795+00:00" ] }, { "id": null, "title": "Interpreting Physics in Video World Models", "source": "https://ai.meta.com/research/publications/interpreting-physics-in-video-world-models/", "recorded_decision": "archive_metadata_only_unconfirmed_nonblocking", "current_computed_status": null, "judgment": "metadata_only_unresolved_identity_or_method_archived_nonblocking_not_implementation_TODO", "full_article_read": false, "review_level": "official_directory_and_frozen_history_identity_check_only", "cost_basis": null, "evidence_paths": [ "docs/research-audits/harness2-priority-repair/meta-audit.json:606" ], "recorded_primary_locations": [ "docs/research-audits/oct15-review-closure/meta-source-review.json: exact URL record", "/tmp/oct20-source-delta/retrieval.json: Meta page 3, checked 2026-10-20T06:08:54.543795+00:00" ] }, { "id": null, "title": "Accurate Decoding of Natural Sentences from Non-Invasive Brain Recordings", "source": "https://ai.meta.com/research/publications/accurate-decoding-of-natural-sentences-from-non-invasive-brain-recordings/", "recorded_decision": "archive_metadata_only_unconfirmed_nonblocking", "current_computed_status": null, "judgment": "metadata_only_unresolved_identity_or_method_archived_nonblocking_not_implementation_TODO", "full_article_read": false, "review_level": "official_directory_and_frozen_history_identity_check_only", "cost_basis": null, "evidence_paths": [ "docs/research-audits/harness2-priority-repair/meta-audit.json:629" ], "recorded_primary_locations": [ "docs/research-audits/oct15-review-closure/meta-source-review.json: exact URL record", "/tmp/oct20-source-delta/retrieval.json: Meta page 3, checked 2026-10-20T06:08:54.543795+00:00" ] }, { "id": "2605.06647", "title": "Superintelligent Retrieval Agent: The Next Frontier of Agentic Retrieval", "source": "https://ai.meta.com/research/publications/superintelligent-retrieval-agent-the-next-frontier-of-agentic-retrieval/", "recorded_decision": "preserve_existing_catalog", "current_computed_status": "implemented", "judgment": "existing_L1_mechanism_diagnostic_not_full_paper_reproduction", "full_article_read": false, "review_level": "catalog_linkage_check", "cost_basis": null, "evidence_paths": [ "docs/research-audits/harness2-priority-repair/meta-audit.json:651", "docs/paper-discovery-ledger.json:13867", "docs/paper-discovery-ledger.json:13872", "docs/research-audits/harness2-priority-repair/backfill-inventory.json:439", "docs/research-manifest.json:3745", "docs/research-manifest.json:3746", "docs/agent-research/2605.06647-sira/README.md:3", "src/auto_research/agent_research/official_meta_backfill_20261004.py:17" ], "recorded_primary_locations": [ "docs/research-audits/oct15-review-closure/meta-source-review.json: exact URL record", "/tmp/oct20-source-delta/retrieval.json: Meta page 3, checked 2026-10-20T06:08:54.543795+00:00" ], "catalog_key": "sira", "detail_path": "docs/agent-research/2605.06647-sira/README.md", "catalog_implementation_status": "absent", "local_metric_file": "docs/agent-research/2605.06647-sira/metrics/mechanism-seeds42-44.json", "local_aggregate_metrics": { "seed_mean": 43, "accepted_terms_mean": 2, "top_document_mean": 1 }, "diagnostic_only": true, "formal_comparison": false }, { "id": "2605.29955", "title": "AutoformBot: Formalizing Mathematics at Scale", "source": "https://ai.meta.com/research/publications/autoformbot-formalizing-mathematics-at-scale/", "recorded_decision": "recommend_reference_catalog_not_reproduced_high_cost", "current_computed_status": "reviewed", "judgment": "terminal_high_cost_review_DONE_not_reproduced_no_automatic_retry", "full_article_read": false, "review_level": "bounded_original_method_evaluation_and_resource_review", "cost_basis": "26 textbooks, thousands of agents; approximately one week per book. Small 39-target ablations use 600M-1200M tokens and hours of runs; full campaign high cost.", "evidence_paths": [ "docs/research-audits/harness2-priority-repair/meta-audit.json:674", "docs/paper-discovery-ledger.json:122642", "docs/paper-discovery-ledger.json:122658", "docs/paper-discovery-ledger.json:122659", "docs/paper-discovery-ledger.json:122666", "docs/paper-discovery-ledger.json:122673", "docs/research-audits/harness2-priority-repair/backfill-inventory.json:466", "docs/research-manifest.json:488", "docs/research-manifest.json:489", "docs/agent-research/2605.29955-autoformbot/README.md:3", "docs/agent-research/2605.29955-autoformbot/README.md:47" ], "recorded_primary_locations": [ "https://arxiv.org/abs/2605.29955: v1 history", "https://arxiv.org/html/2605.29955v1: author affiliations; sections 3-4; section 5.1 week-per-book, 5.2 token-budget ablations", "https://github.com/facebookresearch/autoform-bot: README requirements and MIT license" ], "catalog_key": "autoformbot", "detail_path": "docs/agent-research/2605.29955-autoformbot/README.md", "catalog_implementation_status": "deferred_high_cost" }, { "id": null, "title": "Misalignment Between Backpropagation and the Hierarchy of Brain Responses to Images", "source": "https://ai.meta.com/research/publications/misalignment-between-backpropagation-and-the-hierarchy-of-brain-responses-to-images/", "recorded_decision": "archive_metadata_only_unconfirmed_nonblocking", "current_computed_status": null, "judgment": "metadata_only_unresolved_identity_or_method_archived_nonblocking_not_implementation_TODO", "full_article_read": false, "review_level": "official_directory_and_frozen_history_identity_check_only", "cost_basis": null, "evidence_paths": [ "docs/research-audits/harness2-priority-repair/meta-audit.json:700" ], "recorded_primary_locations": [ "docs/research-audits/oct15-review-closure/meta-source-review.json: exact URL record", "/tmp/oct20-source-delta/retrieval.json: Meta page 3, checked 2026-10-20T06:08:54.543795+00:00" ] }, { "id": null, "title": "EgoBabyVLM: Benchmarking Cross-Modal Learning from Naturalistic Egocentric Video Data", "source": "https://ai.meta.com/research/publications/egobabyvlm-benchmarking-cross-modal-learning-from-naturalistic-egocentric-video-data/", "recorded_decision": "archive_metadata_only_unconfirmed_nonblocking", "current_computed_status": null, "judgment": "metadata_only_unresolved_identity_or_method_archived_nonblocking_not_implementation_TODO", "full_article_read": false, "review_level": "official_directory_and_frozen_history_identity_check_only", "cost_basis": null, "evidence_paths": [ "docs/research-audits/harness2-priority-repair/meta-audit.json:723" ], "recorded_primary_locations": [ "docs/research-audits/oct15-review-closure/meta-source-review.json: exact URL record", "/tmp/oct20-source-delta/retrieval.json: Meta page 3, checked 2026-10-20T06:08:54.543795+00:00" ] }, { "id": null, "title": "GIM: Evaluating models via tasks that integrate multiple cognitive domains", "source": "https://ai.meta.com/research/publications/gim-evaluating-models-via-tasks-that-integrate-multiple-cognitive-domains/", "recorded_decision": "archive_metadata_only_unconfirmed_nonblocking", "current_computed_status": null, "judgment": "metadata_only_unresolved_identity_or_method_archived_nonblocking_not_implementation_TODO", "full_article_read": false, "review_level": "official_directory_and_frozen_history_identity_check_only", "cost_basis": null, "evidence_paths": [ "docs/research-audits/harness2-priority-repair/meta-audit.json:746" ], "recorded_primary_locations": [ "docs/research-audits/oct15-review-closure/meta-source-review.json: exact URL record", "/tmp/oct20-source-delta/retrieval.json: Meta page 3, checked 2026-10-20T06:08:54.543795+00:00" ] }, { "id": null, "title": "NeuralSet: A High-Performing Python Package for Neuro-AI", "source": "https://ai.meta.com/research/publications/neuralset-a-high-performing-python-package-for-neuro-ai/", "recorded_decision": "archive_metadata_only_unconfirmed_nonblocking", "current_computed_status": null, "judgment": "metadata_only_unresolved_identity_or_method_archived_nonblocking_not_implementation_TODO", "full_article_read": false, "review_level": "official_directory_and_frozen_history_identity_check_only", "cost_basis": null, "evidence_paths": [ "docs/research-audits/harness2-priority-repair/meta-audit.json:769" ], "recorded_primary_locations": [ "docs/research-audits/oct15-review-closure/meta-source-review.json: exact URL record", "/tmp/oct20-source-delta/retrieval.json: Meta page 3, checked 2026-10-20T06:08:54.543795+00:00" ] }, { "id": null, "title": "NeuralBench: A Unifying Framework to Benchmark NeuroAI Models", "source": "https://ai.meta.com/research/publications/neuralbench-a-unifying-framework-to-benchmark-neuroai-models/", "recorded_decision": "archive_metadata_only_unconfirmed_nonblocking", "current_computed_status": null, "judgment": "metadata_only_unresolved_identity_or_method_archived_nonblocking_not_implementation_TODO", "full_article_read": false, "review_level": "official_directory_and_frozen_history_identity_check_only", "cost_basis": null, "evidence_paths": [ "docs/research-audits/harness2-priority-repair/meta-audit.json:792" ], "recorded_primary_locations": [ "docs/research-audits/oct15-review-closure/meta-source-review.json: exact URL record", "/tmp/oct20-source-delta/retrieval.json: Meta page 3, checked 2026-10-20T06:08:54.543795+00:00" ] }, { "id": "2605.01188", "title": "Compute Optimal Tokenization", "source": "https://ai.meta.com/research/publications/compute-optimal-tokenization/", "recorded_decision": "recommend_reference_catalog_not_reproduced_high_cost", "current_computed_status": "reviewed", "judgment": "terminal_high_cost_review_DONE_not_reproduced_no_automatic_retry", "full_article_read": false, "review_level": "bounded_original_method_and_resource_review", "cost_basis": "Section 2.2: 988 latent-tokenized plus 320 subword models, 50M-6.7B parameters, 4B-1.1T bytes and 5e18-2e21 FLOPs budgets. Full scaling study not reproduced.", "evidence_paths": [ "docs/research-audits/harness2-priority-repair/meta-audit.json:814", "docs/paper-discovery-ledger.json:122599", "docs/paper-discovery-ledger.json:122615", "docs/paper-discovery-ledger.json:122616", "docs/paper-discovery-ledger.json:122622", "docs/research-audits/harness2-priority-repair/backfill-inventory.json:421" ], "recorded_primary_locations": [ "https://arxiv.org/abs/2605.01188: v1/v2 history", "https://arxiv.org/html/2605.01188v1: affiliation; sections 2.1-2.3 and 3; section2.2 resource grid" ] }, { "id": "2603.26499", "title": "AIRA₂: Overcoming Bottlenecks in AI Research Agents", "source": "https://ai.meta.com/research/publications/aira-overcoming-bottlenecks-in-ai-research-agents/", "recorded_decision": "preserve_existing_catalog", "current_computed_status": "implemented", "judgment": "existing_L1_mechanism_diagnostic_not_full_paper_reproduction", "full_article_read": false, "review_level": "catalog_linkage_check", "cost_basis": null, "evidence_paths": [ "docs/research-audits/harness2-priority-repair/meta-audit.json:838", "docs/paper-discovery-ledger.json:13877", "docs/paper-discovery-ledger.json:13882", "docs/research-audits/harness2-priority-repair/backfill-inventory.json:320", "docs/research-manifest.json:228", "docs/research-manifest.json:229", "docs/agent-research/2603.26499-aira2/README.md:3", "src/auto_research/agent_research/official_meta_backfill_20261004.py:64" ], "recorded_primary_locations": [ "docs/research-audits/oct15-review-closure/meta-source-review.json: exact URL record", "/tmp/oct20-source-delta/retrieval.json: Meta page 3, checked 2026-10-20T06:08:54.543795+00:00" ], "catalog_key": "aira2", "detail_path": "docs/agent-research/2603.26499-aira2/README.md", "catalog_implementation_status": "absent", "local_metric_file": "docs/agent-research/2603.26499-aira2/metrics/mechanism-seeds42-44.json", "local_aggregate_metrics": { "seed_mean": 43, "makespan_mean": 7 }, "diagnostic_only": true, "formal_comparison": false }, { "id": null, "title": "TransText: Transparency Aware Image-to-Video Typography Animation", "source": "https://ai.meta.com/research/publications/transtext-transparency-aware-image-to-video-typography-animation/", "recorded_decision": "archive_metadata_only_unconfirmed_nonblocking", "current_computed_status": null, "judgment": "metadata_only_unresolved_identity_or_method_archived_nonblocking_not_implementation_TODO", "full_article_read": false, "review_level": "official_directory_and_frozen_history_identity_check_only", "cost_basis": null, "evidence_paths": [ "docs/research-audits/harness2-priority-repair/meta-audit.json:862" ], "recorded_primary_locations": [ "docs/research-audits/oct15-review-closure/meta-source-review.json: exact URL record", "/tmp/oct20-source-delta/retrieval.json: Meta page 4, checked 2026-10-20T06:08:55.650641+00:00" ] }, { "id": null, "title": "Think in Strokes, Not Pixels: Process-Driven Image Generation via Interleaved Reasoning", "source": "https://ai.meta.com/research/publications/think-in-strokes-not-pixels-process-driven-image-generation-via-interleaved-reasoning/", "recorded_decision": "archive_metadata_only_unconfirmed_nonblocking", "current_computed_status": null, "judgment": "metadata_only_unresolved_identity_or_method_archived_nonblocking_not_implementation_TODO", "full_article_read": false, "review_level": "official_directory_and_frozen_history_identity_check_only", "cost_basis": null, "evidence_paths": [ "docs/research-audits/harness2-priority-repair/meta-audit.json:885" ], "recorded_primary_locations": [ "docs/research-audits/oct15-review-closure/meta-source-review.json: exact URL record", "/tmp/oct20-source-delta/retrieval.json: Meta page 4, checked 2026-10-20T06:08:55.650641+00:00" ] }, { "id": null, "title": "A foundation model of vision, audition, and language for in-silico neuroscience", "source": "https://ai.meta.com/research/publications/a-foundation-model-of-vision-audition-and-language-for-in-silico-neuroscience/", "recorded_decision": "archive_metadata_only_unconfirmed_nonblocking", "current_computed_status": null, "judgment": "metadata_only_unresolved_identity_or_method_archived_nonblocking_not_implementation_TODO", "full_article_read": false, "review_level": "official_directory_and_frozen_history_identity_check_only", "cost_basis": null, "evidence_paths": [ "docs/research-audits/harness2-priority-repair/meta-audit.json:908" ], "recorded_primary_locations": [ "docs/research-audits/oct15-review-closure/meta-source-review.json: exact URL record", "/tmp/oct20-source-delta/retrieval.json: Meta page 4, checked 2026-10-20T06:08:55.650641+00:00" ] }, { "id": "2603.19461", "title": "HyperAgents", "source": "https://ai.meta.com/research/publications/hyperagents/", "recorded_decision": "preserve_existing_catalog", "current_computed_status": "implemented", "judgment": "existing_L1_mechanism_diagnostic_not_full_paper_reproduction", "full_article_read": false, "review_level": "catalog_linkage_check", "cost_basis": null, "evidence_paths": [ "docs/research-audits/harness2-priority-repair/meta-audit.json:930", "docs/paper-discovery-ledger.json:13897", "docs/paper-discovery-ledger.json:13902", "docs/research-audits/harness2-priority-repair/backfill-inventory.json:302", "docs/research-manifest.json:1887", "docs/research-manifest.json:1888", "docs/agent-research/2603.19461-hyperagents/README.md:3", "src/auto_research/agent_research/official_meta_backfill_20261004.py:156" ], "recorded_primary_locations": [ "docs/research-audits/oct15-review-closure/meta-source-review.json: exact URL record", "/tmp/oct20-source-delta/retrieval.json: Meta page 4, checked 2026-10-20T06:08:55.650641+00:00" ], "catalog_key": "hyperagents", "detail_path": "docs/agent-research/2603.19461-hyperagents/README.md", "catalog_implementation_status": "absent", "local_metric_file": "docs/agent-research/2603.19461-hyperagents/metrics/mechanism-seeds42-44.json", "local_aggregate_metrics": { "seed_mean": 43, "probability_sum_mean": 1, "archive_size_mean": 2 }, "diagnostic_only": true, "formal_comparison": false }, { "id": null, "title": "Omnilingual MT: Machine Translation for 1,600 Languages", "source": "https://ai.meta.com/research/publications/omnilingual-mt-machine-translation-for-1600-languages/", "recorded_decision": "archive_metadata_only_unconfirmed_nonblocking", "current_computed_status": null, "judgment": "metadata_only_unresolved_identity_or_method_archived_nonblocking_not_implementation_TODO", "full_article_read": false, "review_level": "official_directory_and_frozen_history_identity_check_only", "cost_basis": null, "evidence_paths": [ "docs/research-audits/harness2-priority-repair/meta-audit.json:954" ], "recorded_primary_locations": [ "docs/research-audits/oct15-review-closure/meta-source-review.json: exact URL record", "/tmp/oct20-source-delta/retrieval.json: Meta page 4, checked 2026-10-20T06:08:55.650641+00:00" ] }, { "id": null, "title": "Omnilingual SONAR: Cross-Lingual and Cross-Modal Sentence Embeddings Bridging Massively Multilingual Text and Speech", "source": "https://ai.meta.com/research/publications/omnilingual-sonar-cross-lingual-and-cross-modal-sentence-embeddings-bridging-massively-multilingual-text-and-speech/", "recorded_decision": "archive_metadata_only_unconfirmed_nonblocking", "current_computed_status": null, "judgment": "metadata_only_unresolved_identity_or_method_archived_nonblocking_not_implementation_TODO", "full_article_read": false, "review_level": "official_directory_and_frozen_history_identity_check_only", "cost_basis": null, "evidence_paths": [ "docs/research-audits/harness2-priority-repair/meta-audit.json:977" ], "recorded_primary_locations": [ "docs/research-audits/oct15-review-closure/meta-source-review.json: exact URL record", "/tmp/oct20-source-delta/retrieval.json: Meta page 4, checked 2026-10-20T06:08:55.650641+00:00" ] }, { "id": null, "title": "Unified Vision–Language Modeling via Concept Space Alignment", "source": "https://ai.meta.com/research/publications/unified-vision-language-modeling-via-concept-space-alignment/", "recorded_decision": "archive_metadata_only_unconfirmed_nonblocking", "current_computed_status": null, "judgment": "metadata_only_unresolved_identity_or_method_archived_nonblocking_not_implementation_TODO", "full_article_read": false, "review_level": "official_directory_and_frozen_history_identity_check_only", "cost_basis": null, "evidence_paths": [ "docs/research-audits/harness2-priority-repair/meta-audit.json:1000" ], "recorded_primary_locations": [ "docs/research-audits/oct15-review-closure/meta-source-review.json: exact URL record", "/tmp/oct20-source-delta/retrieval.json: Meta page 4, checked 2026-10-20T06:08:55.650641+00:00" ] }, { "id": "2602.16173", "title": "Learning Personalized Agents from Human Feedback", "source": "https://ai.meta.com/research/publications/learning-personalized-agents-from-human-feedback/", "recorded_decision": "preserve_existing_catalog", "current_computed_status": "implemented", "judgment": "existing_L1_mechanism_diagnostic_not_full_paper_reproduction", "full_article_read": false, "review_level": "catalog_linkage_check", "cost_basis": null, "evidence_paths": [ "docs/research-audits/harness2-priority-repair/meta-audit.json:1022", "docs/paper-discovery-ledger.json:13887", "docs/paper-discovery-ledger.json:13892", "docs/research-audits/harness2-priority-repair/backfill-inventory.json:211", "docs/research-manifest.json:2862", "docs/research-manifest.json:2863", "docs/agent-research/2602.16173-pahf/README.md:3", "src/auto_research/agent_research/official_meta_backfill_20261004.py:109" ], "recorded_primary_locations": [ "docs/research-audits/oct15-review-closure/meta-source-review.json: exact URL record", "/tmp/oct20-source-delta/retrieval.json: Meta page 4, checked 2026-10-20T06:08:55.650641+00:00" ], "catalog_key": "pahf", "detail_path": "docs/agent-research/2602.16173-pahf/README.md", "catalog_implementation_status": "absent", "local_metric_file": "docs/agent-research/2602.16173-pahf/metrics/mechanism-seeds42-44.json", "local_aggregate_metrics": { "seed_mean": 43 }, "diagnostic_only": true, "formal_comparison": false }, { "id": null, "title": "FERRET: Framework for Expansion Reliant Red Teaming", "source": "https://ai.meta.com/research/publications/ferret-framework-for-expansion-reliant-red-teaming/", "recorded_decision": "archive_metadata_only_unconfirmed_nonblocking", "current_computed_status": null, "judgment": "metadata_only_unresolved_identity_or_method_archived_nonblocking_not_implementation_TODO", "full_article_read": false, "review_level": "official_directory_and_frozen_history_identity_check_only", "cost_basis": null, "evidence_paths": [ "docs/research-audits/harness2-priority-repair/meta-audit.json:1046" ], "recorded_primary_locations": [ "docs/research-audits/oct15-review-closure/meta-source-review.json: exact URL record", "/tmp/oct20-source-delta/retrieval.json: Meta page 4, checked 2026-10-20T06:08:55.650641+00:00" ] }, { "id": null, "title": "UniT: Unified Multimodal Chain-of-Thought Test-time Scaling", "source": "https://ai.meta.com/research/publications/unit-unified-multimodal-chain-of-thought-test-time-scaling/", "recorded_decision": "archive_metadata_only_unconfirmed_nonblocking", "current_computed_status": null, "judgment": "metadata_only_unresolved_identity_or_method_archived_nonblocking_not_implementation_TODO", "full_article_read": false, "review_level": "official_directory_and_frozen_history_identity_check_only", "cost_basis": null, "evidence_paths": [ "docs/research-audits/harness2-priority-repair/meta-audit.json:1069" ], "recorded_primary_locations": [ "docs/research-audits/oct15-review-closure/meta-source-review.json: exact URL record", "/tmp/oct20-source-delta/retrieval.json: Meta page 4, checked 2026-10-20T06:08:55.650641+00:00" ] }, { "id": "2602.06855", "title": "AIRS-Bench: a Suite of Tasks for Frontier AI Research Science Agents", "source": "https://ai.meta.com/research/publications/airs-bench-a-suite-of-tasks-for-frontier-ai-research-science-agents/", "recorded_decision": "recommend_reference_catalog_not_reproduced_high_cost; affordable task-metadata/result-import interface possible, not currently integrated", "current_computed_status": "reviewed", "judgment": "terminal_high_cost_review_DONE_not_reproduced_no_automatic_retry", "full_article_read": false, "review_level": "bounded_original_protocol_and_resource_review", "cost_basis": "20 tasks, each research run 24h on one H200; Appendix D gives 10-20 runs/task. A single original run already exceeds cheap budget; complete campaign not reproduced.", "evidence_paths": [ "docs/research-audits/harness2-priority-repair/meta-audit.json:1091", "docs/paper-discovery-ledger.json:122413", "docs/paper-discovery-ledger.json:122429", "docs/paper-discovery-ledger.json:122430", "docs/paper-discovery-ledger.json:122437", "docs/paper-discovery-ledger.json:122444", "docs/research-audits/harness2-priority-repair/backfill-inventory.json:175", "docs/research-manifest.json:246", "docs/research-manifest.json:247", "docs/agent-research/2602.06855-airs-bench/README.md:3", "docs/agent-research/2602.06855-airs-bench/README.md:46" ], "recorded_primary_locations": [ "https://arxiv.org/abs/2602.06855: version history", "https://arxiv.org/html/2602.06855v1: sections 4.2-4.3 and 5.1; Appendix C Table7, Appendix D", "https://github.com/facebookresearch/airs-bench: README task specification and CC BY-NC 4.0 license" ], "catalog_key": "airs-bench", "detail_path": "docs/agent-research/2602.06855-airs-bench/README.md", "catalog_implementation_status": "deferred_high_cost" }, { "id": null, "title": "PhyGDPO: Physics-Aware Groupwise Direct Preference Optimization for Physically Consistent Text-to-Video Generation", "source": "https://ai.meta.com/research/publications/phygdpo-physics-aware-groupwise-direct-preference-optimization-for-physically-consistent-text-to-video-generation/", "recorded_decision": "archive_metadata_only_unconfirmed_nonblocking", "current_computed_status": null, "judgment": "metadata_only_unresolved_identity_or_method_archived_nonblocking_not_implementation_TODO", "full_article_read": false, "review_level": "official_directory_and_frozen_history_identity_check_only", "cost_basis": null, "evidence_paths": [ "docs/research-audits/harness2-priority-repair/meta-audit.json:1117" ], "recorded_primary_locations": [ "docs/research-audits/oct15-review-closure/meta-source-review.json: exact URL record", "/tmp/oct20-source-delta/retrieval.json: Meta page 4, checked 2026-10-20T06:08:55.650641+00:00" ] }, { "id": "2610.07338", "title": "Logbook: Extremely Long-form Audio Event Understanding", "source": "https://arxiv.org/html/2610.07338v1", "recorded_decision": "not_reproduced_high_cost", "current_computed_status": "reviewed", "judgment": "terminal_high_cost_review_DONE_not_reproduced_no_automatic_retry", "full_article_read": false, "review_level": "bounded_original_resource_gate_check_with_prior_method_review_preserved", "cost_basis": "52 systems; Table1 test audio82.1h+240.6h+148.9h, full cascaded/LLM fact+entailment evaluation. Full campaign high cost.", "evidence_paths": [ "docs/research-audits/harness2-priority-repair/meta-audit.json:1141", "docs/paper-discovery-ledger.json:19162", "docs/paper-discovery-ledger.json:19176", "docs/paper-discovery-ledger.json:122868", "docs/paper-discovery-ledger.json:122886", "docs/paper-discovery-ledger.json:122887", "docs/paper-discovery-ledger.json:122892", "docs/paper-discovery-ledger.json:122893", "docs/paper-discovery-ledger.json:122907", "docs/research-audits/harness2-priority-repair/backfill-inventory.json:681" ], "recorded_primary_locations": [ "docs/paper-discovery-ledger.json: exact id in old deferred record", "docs/research-audits/oct15-review-closure/implementation-status.json: review_decision known, current_status not_catalogued, empty observations", "https://arxiv.org/html/2610.07338v1: sections2.2-2.4 Table1", "https://arxiv.org/html/2610.07338v1: sections3-5 systems/evaluation" ] }, { "id": "2610.05417", "title": "Render to Reason: Novel-View Semantic Prediction Improves Spatial Understanding in VLMs", "source": "https://arxiv.org/html/2610.05417v1", "recorded_decision": "not_reproduced_high_cost", "current_computed_status": "reviewed", "judgment": "terminal_high_cost_review_DONE_not_reproduced_no_automatic_retry", "full_article_read": false, "review_level": "bounded_original_resource_gate_check_with_prior_method_review_preserved", "cost_basis": "AppendixA.1:8GH200;4B variants12h,InternVL8B30h,Qwen3VL8B96h per training configuration, plus ablations.", "evidence_paths": [ "docs/research-audits/harness2-priority-repair/meta-audit.json:1172", "docs/paper-discovery-ledger.json:24895", "docs/paper-discovery-ledger.json:24909", "docs/paper-discovery-ledger.json:122824", "docs/paper-discovery-ledger.json:122842", "docs/paper-discovery-ledger.json:122847", "docs/paper-discovery-ledger.json:122848", "docs/paper-discovery-ledger.json:122862", "docs/research-audits/harness2-priority-repair/backfill-inventory.json:672" ], "recorded_primary_locations": [ "docs/paper-discovery-ledger.json: exact id in old deferred record", "docs/research-audits/oct15-review-closure/implementation-status.json: review_decision known, current_status not_catalogued, empty observations", "https://arxiv.org/html/2610.05417v1: AppendixA.1 training configuration and wall clock" ] }, { "id": "2610.09015", "title": "Personalize at Test Time: Learning User Preferences for Image Generation", "source": "https://arxiv.org/html/2610.09015v1", "recorded_decision": "not_reproduced_high_cost", "current_computed_status": "reviewed", "judgment": "terminal_high_cost_review_DONE_not_reproduced_no_automatic_retry", "full_article_read": false, "review_level": "bounded_original_resource_gate_check_with_prior_method_review_preserved", "cost_basis": "50000 images annotated by72B VLM,36000 evaluator pairs;631users x10images/method,30 diffusion sampling steps x30 DNO updates plus GPT5.5judging.", "evidence_paths": [ "docs/research-audits/harness2-priority-repair/meta-audit.json:1202", "docs/paper-discovery-ledger.json:42646", "docs/paper-discovery-ledger.json:42659", "docs/paper-discovery-ledger.json:42669", "docs/paper-discovery-ledger.json:122913", "docs/paper-discovery-ledger.json:122931", "docs/paper-discovery-ledger.json:122932", "docs/paper-discovery-ledger.json:122933", "docs/paper-discovery-ledger.json:122938", "docs/research-audits/harness2-priority-repair/backfill-inventory.json:690" ], "recorded_primary_locations": [ "docs/paper-discovery-ledger.json: exact id in old deferred record", "docs/research-audits/oct15-review-closure/implementation-status.json: review_decision known, current_status not_catalogued, empty observations", "https://arxiv.org/html/2610.09015v1: section3.2 evaluator annotation/training", "https://arxiv.org/html/2610.09015v1: section4.4 personalized generation", "https://arxiv.org/html/2610.09015v1: AppendixD DNO" ] }, { "id": "2610.09202", "title": "Few Bits, One Law: Toward W2A4KV2", "source": "https://arxiv.org/html/2610.09202v1", "recorded_decision": "not_reproduced_high_cost", "current_computed_status": "reviewed", "judgment": "terminal_high_cost_review_DONE_not_reproduced_no_automatic_retry", "full_article_read": false, "review_level": "bounded_original_resource_gate_check_with_prior_method_review_preserved", "cost_basis": "AppendixA.1:8H100/A10080GB,up to240000QATupdates/about31.5Btokens plus8K40KcontinuedQAT.", "evidence_paths": [ "docs/research-audits/harness2-priority-repair/meta-audit.json:1234", "docs/paper-discovery-ledger.json:43517", "docs/paper-discovery-ledger.json:43530", "docs/paper-discovery-ledger.json:43540", "docs/paper-discovery-ledger.json:122959", "docs/paper-discovery-ledger.json:122977", "docs/paper-discovery-ledger.json:122978", "docs/paper-discovery-ledger.json:122983", "docs/paper-discovery-ledger.json:122984", "docs/research-audits/harness2-priority-repair/backfill-inventory.json:699" ], "recorded_primary_locations": [ "docs/paper-discovery-ledger.json: exact id in old deferred record", "docs/research-audits/oct15-review-closure/implementation-status.json: review_decision known, current_status not_catalogued, empty observations", "https://arxiv.org/html/2610.09202v1: section5.1 evaluation setup", "https://arxiv.org/html/2610.09202v1: AppendixA.1 optimization" ] }, { "id": "2610.09440", "title": "Mixture of Layers: Dynamic Layer Routing for Visual Reasoning", "source": "https://arxiv.org/html/2610.09440v1", "recorded_decision": "not_reproduced_high_cost", "current_computed_status": "reviewed", "judgment": "terminal_high_cost_review_DONE_not_reproduced_no_automatic_retry", "full_article_read": false, "review_level": "bounded_original_resource_gate_check_with_prior_method_review_preserved", "cost_basis": "AppendixB:4H100NVL training,558K+665K LLaVA or2M Bunny data;13B/8B backbones and multiple routers/ablations.", "evidence_paths": [ "docs/research-audits/harness2-priority-repair/meta-audit.json:1265", "docs/paper-discovery-ledger.json:44717", "docs/paper-discovery-ledger.json:44730", "docs/paper-discovery-ledger.json:44740", "docs/paper-discovery-ledger.json:123004", "docs/paper-discovery-ledger.json:123022", "docs/paper-discovery-ledger.json:123027", "docs/paper-discovery-ledger.json:123028", "docs/paper-discovery-ledger.json:123042", "docs/research-audits/harness2-priority-repair/backfill-inventory.json:708" ], "recorded_primary_locations": [ "docs/paper-discovery-ledger.json: exact id in old deferred record", "docs/research-audits/oct15-review-closure/implementation-status.json: review_decision known, current_status not_catalogued, empty observations", "https://arxiv.org/html/2610.09440v1: AppendixB hardware and Table7 training recipes" ] }, { "id": "2610.10990", "title": "Omni-Diffusion-Distill: Few-Step Distillation of Unified Multimodal Diffusion Large Language Models", "source": "https://arxiv.org/html/2610.10990v1", "recorded_decision": "not_reproduced_high_cost", "current_computed_status": "reviewed", "judgment": "terminal_high_cost_review_DONE_not_reproduced_no_automatic_retry", "full_article_read": false, "review_level": "bounded_original_resource_gate_check_with_prior_method_review_preserved", "cost_basis": "AppendixC.2 Table7:32A100,total110.8h/3546GPU-hours including teacher collection and2trainingstages.", "evidence_paths": [ "docs/research-audits/harness2-priority-repair/meta-audit.json:1295", "docs/paper-discovery-ledger.json:51304", "docs/paper-discovery-ledger.json:51317", "docs/paper-discovery-ledger.json:51327", "docs/paper-discovery-ledger.json:123048", "docs/paper-discovery-ledger.json:123066", "docs/paper-discovery-ledger.json:123071", "docs/paper-discovery-ledger.json:123072", "docs/paper-discovery-ledger.json:123086", "docs/research-audits/harness2-priority-repair/backfill-inventory.json:717" ], "recorded_primary_locations": [ "docs/paper-discovery-ledger.json: exact id in old deferred record", "docs/research-audits/oct15-review-closure/implementation-status.json: review_decision known, current_status not_catalogued, empty observations", "https://arxiv.org/html/2610.10990v1: AppendixC.2 Tables7-8" ] }, { "id": "2610.11019", "title": "Mid-Training Language Models on Raw Video", "source": "https://arxiv.org/html/2610.11019v1", "recorded_decision": "recommend_reference_catalog_not_reproduced_high_cost", "current_computed_status": "reviewed", "judgment": "terminal_high_cost_review_DONE_not_reproduced_no_automatic_retry", "full_article_read": false, "review_level": null, "cost_basis": "Original section4.1.4: end-to-end27300updates over128H100, batch256,16kcontext, plus matched instruction tuning. No download or run.", "evidence_paths": [ "docs/research-audits/harness2-priority-repair/meta-audit.json:1325", "docs/paper-discovery-ledger.json:51495", "docs/paper-discovery-ledger.json:51508", "docs/paper-discovery-ledger.json:51518", "docs/paper-discovery-ledger.json:123092", "docs/paper-discovery-ledger.json:123108", "docs/paper-discovery-ledger.json:123109", "docs/paper-discovery-ledger.json:123116", "docs/paper-discovery-ledger.json:123117", "docs/research-audits/harness2-priority-repair/backfill-inventory.json:726", "docs/research-manifest.json:11611", "docs/research-manifest.json:11612", "docs/multimodal-models/2610.11019-raw-video-midtraining/README.md:3", "docs/multimodal-models/2610.11019-raw-video-midtraining/README.md:48" ], "recorded_primary_locations": [ "https://arxiv.org/abs/2610.11019: v1 history", "https://arxiv.org/html/2610.11019v1: affiliations; section3 next-token objective; section4.1.4 training", "https://jd730.github.io/projects/RawVideoMidTraining/: project source links", "docs/research-audits/oct15-review-closure/implementation-status.json: not_catalogued" ], "catalog_key": "raw-video-midtraining", "detail_path": "docs/multimodal-models/2610.11019-raw-video-midtraining/README.md", "catalog_implementation_status": "deferred_high_cost" }, { "id": "2610.11316", "title": "MetaEncoder: Exploring the Limit of Bi-Encoders for Multimodal System One Decision Making with Natural Language Interface", "source": "https://arxiv.org/html/2610.11316v1", "recorded_decision": "recommend_reference_catalog_not_reproduced_high_cost", "current_computed_status": "reviewed", "judgment": "terminal_high_cost_review_DONE_not_reproduced_no_automatic_retry", "full_article_read": false, "review_level": null, "cost_basis": "Official eval README: 30B model about62GB bf16 per GPU, intended8x80GB; several hundredGB MMEB-V2 data,190tasks. No download or run.", "evidence_paths": [ "docs/research-audits/harness2-priority-repair/meta-audit.json:1354", "docs/paper-discovery-ledger.json:53026", "docs/paper-discovery-ledger.json:53039", "docs/paper-discovery-ledger.json:123137", "docs/paper-discovery-ledger.json:123153", "docs/paper-discovery-ledger.json:123154", "docs/paper-discovery-ledger.json:123161", "docs/paper-discovery-ledger.json:123162", "docs/paper-discovery-ledger.json:123169", "docs/research-audits/harness2-priority-repair/backfill-inventory.json:735", "docs/research-manifest.json:7935", "docs/research-manifest.json:7936", "docs/foundation-models/2610.11316-metaencoder/README.md:3", "docs/foundation-models/2610.11316-metaencoder/README.md:45" ], "recorded_primary_locations": [ "https://arxiv.org/abs/2610.11316: v1 history", "https://arxiv.org/html/2610.11316v1: author block; sections3.1-3.3 and4", "https://github.com/facebook/meta-encoder-eval: README setup, resource and data requirements", "docs/research-audits/oct15-review-closure/implementation-status.json: not_catalogued" ], "catalog_key": "metaencoder", "detail_path": "docs/foundation-models/2610.11316-metaencoder/README.md", "catalog_implementation_status": "deferred_high_cost" }, { "id": "2610.13886", "title": "X-AgentRed: Red-Teaming Personal Agents in Evolving Multi-Service Ecosystems", "source": "https://arxiv.org/html/2610.13886v1", "recorded_decision": "not_reproduced_high_cost", "current_computed_status": "reviewed", "judgment": "terminal_high_cost_review_DONE_not_reproduced_no_automatic_retry", "full_article_read": false, "review_level": "bounded_original_resource_gate_check_with_prior_method_review_preserved", "cost_basis": "4frontier target agents x250scenarios,30services914tools; Opus4.8simulator/redteam,dualjudge;22.10generationAPIcalls/scenario excludes execution/judging; costs multiply for baselines.", "evidence_paths": [ "docs/research-audits/harness2-priority-repair/meta-audit.json:1383", "docs/paper-discovery-ledger.json:62088", "docs/paper-discovery-ledger.json:123182", "docs/paper-discovery-ledger.json:123200", "docs/paper-discovery-ledger.json:123201", "docs/paper-discovery-ledger.json:123206", "docs/paper-discovery-ledger.json:123207", "docs/paper-discovery-ledger.json:123221", "docs/research-audits/harness2-priority-repair/backfill-inventory.json:744" ], "recorded_primary_locations": [ "docs/paper-discovery-ledger.json: exact id in old deferred record", "docs/research-audits/oct15-review-closure/implementation-status.json: review_decision known, current_status not_catalogued, empty observations", "https://arxiv.org/html/2610.13886v1: sections4.1 and4.6", "https://arxiv.org/html/2610.13886v1: AppendixB.4 generation cost accounting" ] } ], "new_original_checks": [ { "id": "2610.06064", "source": "https://arxiv.org/html/2610.06064v1", "locations": "§4; Appendix C; Appendix D.2-D.4 (web lines 164-170, 763-802)" }, { "id": "2610.04721", "source": "https://arxiv.org/html/2610.04721v1", "locations": "Appendix E.1 Table10 (web lines 681-693)" }, { "id": "2610.09484", "source": "https://arxiv.org/html/2610.09484v1", "locations": "Appendix A.1 Table4, A.3 Table7, D.1 Table13 (web lines 304-328, 382-427, 658-678)" } ], "no_repository_changes": true, "no_tests_or_experiments_run": true }