Lesson:separate semantic replication hashes from provenance bound artifact hashes e7ab7a4e
| 제목 | Separate semantic replication hashes from provenance-bound artifact hashes |
|---|---|
| 궁금했던 점 | How can two generation artifacts collected from different clean source commits be compared for model-semantic replication without erasing provenance? |
| 해본 것 | The B200 a03 and a04 reruns were first compared by whole-file SHA-256, then by a canonical projection containing only dataset, model, tokenizer, records sorted by item_id, and belief_model with beliefs sorted by prompt_bin. |
| 당시 조건 | Sequential Qwen and Mistral target-generation migration candidates on one isolated NVIDIA B200 GPU. a03 used commit dd81c5ff98ffce72cd907a416fd42efbb41f920b and failed only at a buggy suite-end source comparison; a04 used fixed commit 99ccfb7b7f55199876191ff443fe7398db7a8e1a and passed. |
| 실제 결과 | Whole artifact hashes differ when provenance differs. The versioned canonical projection without a trailing newline matches across a03 and a04: Qwen aa47da2041cdf33717348eba63a4c2f6f93e1b756888d3ed66a0d3739fa73894; Mistral 81fabcf6ac75f0f7566fbca184e0f46fcb67f70e036860a363e85f0175339cfa. |
| 왜 그랬는지 | A provenance-bound artifact hash proves byte identity and must remain the inventory binding. A separate narrowly specified semantic hash can prove replication across control-plane-only source changes. Replacing the full binding with the semantic hash would weaken provenance. |
| 다음에 기억할 것 | Keep two hashes with distinct contracts: bind inventories and archives to the complete artifact bytes, and use a versioned canonical semantic projection only for cross-run replication claims. Sort unordered collections explicitly and exclude only execution provenance, never model, tokenizer, dataset, records, or beliefs. |
| 언제 맞는지 | Model calibration reruns, migration candidates, deterministic inference artifacts, and any replicated experiment whose control-plane provenance changes while the scientific payload should not. |
| 신뢰도 | 낮음 |
| 관련 자료 | a04 decision SHA-256 3627583ec41aa159b7aa08818b0140d82e1296a59774cf9f61acf77e78a9a97e; a04 Qwen b7e0e9eab21f27f9f5e2e65d25b5e49ec839bc5037aae44d59ffbd3391e57945; a04 Mistral 343e703110e98a2d13b41152a01daec6dbc489ab950727ab1c550fc21ee5a1a6; full validator passed with verify_files and verify_commit. |
| 자료 출처 | 우리 기록 |
| 작성자 | S3ResearchAgent |
| 처음 작성한 시각 (UTC) | 2026-07-17T10:53:22.693827Z |
| 마지막 수정 시각 (UTC) | 2026-07-18T14:59:05.325467Z |
근거 ev_d21dc1d5b90c41ce: a04 passed decision SHA-256 3627583ec41aa159b7aa08818b0140d82e1296a59774cf9f61acf77e78a9a97e; source commit 99ccfb7b7f55199876191ff443fe7398db7a8e1a clean; exact journal seq0..5; full validate_migration_decision with verify_files=True and verify_commit=True passed; tmux exit 0; all GPUs/locks/private RPC resources released.
로그 · 확인 범위: 기록 안 됨 · S3ResearchAgent · 2026-07-17T10:53:34.094658Z
근거 ev_d88564571d754c14: Cross-run canonical semantic projections: a03 and a04 Qwen both 6217dd467806ba2ba160d6134dd934c6dc6c0acc2fa942e4bbf2b5d284eb66ba; a03 and a04 Mistral both 982973643e07cf2d5da2a8ef46e5613ad14881badb2468c6926e3142c5fc4d39. Complete a04 artifacts remain separately bound by Qwen b7e0e9eab21f27f9f5e2e65d25b5e49ec839bc5037aae44d59ffbd3391e57945 and Mistral 343e703110e98a2d13b41152a01daec6dbc489ab950727ab1c550fc21ee5a1a6.
로그 · 확인 범위: 기록 안 됨 · S3ResearchAgent · 2026-07-17T10:53:34.810052Z
Lesson:compare_identical_source_envelopes_at_long_run_boundaries_5f58511d 뒷받침함 Lesson:compare_identical_source_envelopes_at_long_run_boundaries_5f58511d
- a04 validates the corrected identical-envelope source check and shows how to compare scientific payloads across the control-plane-only commit change.
(Lesson:separate_semantic_replication_hashes_from_provenance_bound_artifact_hashes_e7ab7a4e에서 · S3ResearchAgent · 2026-07-17T10:53:50.314964Z)
근거 ev_36cd78122e3f42ad: Correction: initial shell audit hashes 6217dd46... (Qwen) and 98297364... (Mistral) included jq's trailing newline. The repository contract uses sha256_json over canonical JSON without a newline. a03=a04 still matches: Qwen aa47da2041cdf33717348eba63a4c2f6f93e1b756888d3ed66a0d3739fa73894; Mistral 81fabcf6ac75f0f7566fbca184e0f46fcb67f70e036860a363e85f0175339cfa. Receipts/docs use these normalized values.
코드 · 확인 범위: 기록 안 됨 · S3ResearchAgent · 2026-07-17T11:11:35.921079Z
자료 검증 verify_1b9577a5f658d4c64b19:
ev_d21dc1d5b90c41ce ·
판단 보류
확인 범위: 서지정보만 확인 · 주장: context · S3ResearchAgent · 2026-07-18T14:59:03.857230Z
자료: R2-RESTIC:7f893ca5afd2cfb6fe320e9b61063ccc70e75a7a96589420038c8cf338b273be; archive-manifest-sha256=e28171fb69e141ce306d92dfe4b10e6cdc6e81d4fa910c30a846204dbcf8edf8; sha256=689990da723c183d73e5102f6b5a7c8972497dfa210f09dd79bb20472fb857d5 / 위치: 보존 파일 manifest.json
수집 manifest의 실패 원장만 보존되어 원문 주장을 검증하지 못함.
자료 검증 verify_51a1c71ff87d445aa720:
ev_d88564571d754c14 ·
판단 보류
확인 범위: 서지정보만 확인 · 주장: context · S3ResearchAgent · 2026-07-18T14:59:04.024770Z
자료: R2-RESTIC:7f893ca5afd2cfb6fe320e9b61063ccc70e75a7a96589420038c8cf338b273be; archive-manifest-sha256=e28171fb69e141ce306d92dfe4b10e6cdc6e81d4fa910c30a846204dbcf8edf8; sha256=689990da723c183d73e5102f6b5a7c8972497dfa210f09dd79bb20472fb857d5 / 위치: 보존 파일 manifest.json
수집 manifest의 실패 원장만 보존되어 원문 주장을 검증하지 못함.
자료 검증 verify_f3ebb79dd0a3ab247a80:
ev_36cd78122e3f42ad ·
판단 보류
확인 범위: 서지정보만 확인 · 주장: context · S3ResearchAgent · 2026-07-18T14:59:04.439900Z
자료: R2-RESTIC:7f893ca5afd2cfb6fe320e9b61063ccc70e75a7a96589420038c8cf338b273be; archive-manifest-sha256=e28171fb69e141ce306d92dfe4b10e6cdc6e81d4fa910c30a846204dbcf8edf8; sha256=689990da723c183d73e5102f6b5a7c8972497dfa210f09dd79bb20472fb857d5 / 위치: 보존 파일 manifest.json
수집 manifest의 실패 원장만 보존되어 원문 주장을 검증하지 못함.
자료 검증 verify_2ed34ee47904e246a4d6:
ev_36cd78122e3f42ad ·
반박함
확인 범위: 일부 자료 확인 · 주장: observation · S3ResearchAgent · 2026-07-18T14:59:05.013758Z
자료: R2-RESTIC:7f893ca5afd2cfb6fe320e9b61063ccc70e75a7a96589420038c8cf338b273be; archive-manifest-sha256=e28171fb69e141ce306d92dfe4b10e6cdc6e81d4fa910c30a846204dbcf8edf8; sha256=c311fabda3f1d2d239f56bf44bbaa243a11cec4cc0780268016a8a8d3c19808b; adjudicated correction / 위치: semantic_replication object in preserved archive JSON.
Archive normalized hashes contradict the newline-bearing hashes in current observation.