{ "datasets": [ { "contamination_status": "fixed-26 exact task-ID overlap is zero; public benchmark-family semantic overlap remains unresolved", "dataset_id": "sft-v1-321", "excluded_from_hf": [ "The unused 40-row validation and 40-row test projections and their split_manifest are not preserved." ], "files": { "data/consumption_receipt.json": { "sha256": "d229b8e5fbfe4ea77e2cde8f3f5061a4ff8a75dd0e4bd774ea1cd819fc99d749", "size_bytes": 3367 }, "data/manifest.json": { "sha256": "392d7e6656cd5da48d15a6a035fe42ed89cf09eae30ec25bbb7205b09b01e39c", "size_bytes": 931 }, "data/sft/source-401.jsonl": { "sha256": "2ddfe6966c828007f7d6c439e51bfaf07c8959dddc4423ceadb602dc3d49517b", "size_bytes": 2235935 }, "data/sft/train.jsonl": { "sha256": "257f2aef6ccb1d3f02764eeb976c0ea233f2f36fb5b2dbd26ac8b1d4f22037a5", "size_bytes": 1872928 }, "data/source/train-401-dsa.zip": { "sha256": "2efe714c454de7ba1c5fd523f5849b3c6c9af65e8e5ca5bf4cf438017fb1e03a", "size_bytes": 131776 } }, "fixed26_task_id_overlap": 0, "kind": "glm47-aider-dataset-entry", "license_status": "unresolved-private-preservation-only", "lineage": "Deterministic 321/40/40 split of the original 401-row source", "oracle_visibility": "internal review or verification receipts may be present; no hidden benchmark/runtime oracle", "role": "Exact SFT v1 active split and its 401-row source lineage", "rows": "401 source; 321 train; 320 consumed", "schema_version": 1, "split_semantics": "321-row active train split; seed-42 shuffle consumed ten complete 32-row batches and omitted active-split index 57", "status": "historical-trained", "trainable": false }, { "contamination_status": "exact task-id clear for canonical train/validation/monitor payloads; semantic benchmark-family contamination is unresolved unless the entry audit says otherwise", "dataset_id": "sft-v2-1211", "excluded_from_hf": [ "Raw Training_jsonl.zip sources are provenance-only and not republished." ], "files": { "data/manifest.json": { "sha256": "c2a412ba0f9029b5383ce99dadd56f8e93b0d3608bf2a9baf15e4e2b1140ed67", "size_bytes": 1361 }, "data/sft/train.jsonl": { "sha256": "13219cae85551714d4280b60600bb7ef5336dffda54698340ba40f3405ccd51b", "size_bytes": 5808888 }, "evidence.tar.gz": { "sha256": "62df8e1711371e1bf28442afbd44c0ff43937f607084ec036ab83c42f9745c26", "size_bytes": 646 } }, "fixed26_task_id_overlap": 0, "kind": "glm47-aider-dataset-entry", "license_status": "unresolved-private-preservation-only", "lineage": "401 curated + 480 DSA + 330 puzzle rows", "oracle_visibility": "internal review or verification receipts may be present; no hidden benchmark/runtime oracle", "role": "SFT training corpus", "rows": "1,211 packaged; 1,184 consumed", "schema_version": 1, "split_semantics": "not-applicable", "status": "historical-trained", "trainable": false }, { "contamination_status": "exact task-id clear for canonical train/validation/monitor payloads; semantic benchmark-family contamination is unresolved unless the entry audit says otherwise", "dataset_id": "curated-heuristic-v1", "excluded_from_hf": [], "files": { "data/manifest.json": { "sha256": "72a619f9ee2b2bb8c0ec852cf5510913b882b42e3a48e284bcb0cf6debcf1d41", "size_bytes": 628 }, "data/sft/train.jsonl": { "sha256": "670934a57e1ecdf9f5c930801a235e8e097b756f669615461073ae8420c8b29d", "size_bytes": 174576 }, "evidence.tar.gz": { "sha256": "7c3a3d082a812754d9b6c2d5330be86e306495adc54cf0b8a7b433e9c118d211", "size_bytes": 53067 } }, "fixed26_task_id_overlap": 0, "kind": "glm47-aider-dataset-entry", "license_status": "unresolved-private-preservation-only", "lineage": "Heuristic selection from SFT v2", "oracle_visibility": "internal review or verification receipts may be present; no hidden benchmark/runtime oracle", "role": "Candidate SFT selection", "rows": "32 rows", "schema_version": 1, "split_semantics": "not-applicable", "status": "review-only-not-train-ready", "trainable": false }, { "contamination_status": "exact task-id clear for canonical train/validation/monitor payloads; semantic benchmark-family contamination is unresolved unless the entry audit says otherwise", "dataset_id": "curated-high-confidence-v1", "excluded_from_hf": [], "files": { "data/manifest.json": { "sha256": "018ca5e1656bc27ba3da9049631df829260cac8b71aceca2dd791d67f62579a0", "size_bytes": 780 }, "data/sft/train.jsonl": { "sha256": "6744ec0d10602175b1ba2f1b7a01ef06a754ebb05e7ab0ab1eac282ef6168aba", "size_bytes": 70692 }, "evidence.tar.gz": { "sha256": "9fd60061641f44aea041732fa94417d25a065ede3a66d7be06cf9d4c58579875", "size_bytes": 58971 } }, "fixed26_task_id_overlap": 0, "kind": "glm47-aider-dataset-entry", "license_status": "unresolved-private-preservation-only", "lineage": "High-confidence subset of curated-heuristic-v1", "oracle_visibility": "internal review or verification receipts may be present; no hidden benchmark/runtime oracle", "role": "Candidate SFT selection", "rows": "14 rows", "schema_version": 1, "split_semantics": "not-applicable", "status": "review-only-not-train-ready", "trainable": false }, { "contamination_status": "exact task-id clear for canonical train/validation/monitor payloads; semantic benchmark-family contamination is unresolved unless the entry audit says otherwise", "dataset_id": "sft-gold-v4", "excluded_from_hf": [], "files": { "data/manifest.json": { "sha256": "13b8b84440b7711b4b66d079f2baac66a8d97621693f3f2739846149937fe682", "size_bytes": 2353 }, "data/sft/train.jsonl": { "sha256": "465d1cc7fcf35be91204a0903e2625abd439eaf396c0c8d85487397c2d2cf223", "size_bytes": 1151887 }, "data/validation/validation.jsonl": { "sha256": "488fe3aedae53157f22b046d4ff9c46b9d3f6741f36bf59cfc50f894609780d8", "size_bytes": 329313 }, "evidence.tar.gz": { "sha256": "6a448b8bce324c3f493c07f03e12ee2f3f1431ad47b60dcdfdb393d55c45d9c0", "size_bytes": 7316 } }, "fixed26_task_id_overlap": 0, "kind": "glm47-aider-dataset-entry", "license_status": "unresolved-private-preservation-only", "lineage": "75-source rows plus 85 verified signal-v3 rows", "oracle_visibility": "internal review or verification receipts may be present; no hidden benchmark/runtime oracle", "role": "Verified SFT candidate", "rows": "160 train + 22 validation", "schema_version": 1, "split_semantics": "not-applicable", "status": "evaluated-candidate", "trainable": false }, { "contamination_status": "exact task-id clear for canonical train/validation/monitor payloads; semantic benchmark-family contamination is unresolved unless the entry audit says otherwise", "dataset_id": "raw-concat-v1", "excluded_from_hf": [], "files": { "data/manifest.json": { "sha256": "9116cb2edac699c95ed69f96fd042995da8e3902e89b026f97ae80d0d102c7c9", "size_bytes": 851 }, "data/sft/train.jsonl": { "sha256": "e1b6b8328f5104f8b559d9b5edbc21baf0aec54ead83fcf0e32bb42baaef7d98", "size_bytes": 5035550 }, "evidence.tar.gz": { "sha256": "78cd633b2f1a11dc39713ef116372f1dd89a2fc4584be4f32b83c081c9ae0923", "size_bytes": 16194 } }, "fixed26_task_id_overlap": 0, "kind": "glm47-aider-dataset-entry", "license_status": "unresolved-private-preservation-only", "lineage": "160 Gold-v4 + 715 reverify rows; 7 duplicate task IDs", "oracle_visibility": "internal review or verification receipts may be present; no hidden benchmark/runtime oracle", "role": "Literal concatenation SFT corpus", "rows": "875 rows; 868 unique task IDs", "schema_version": 1, "split_semantics": "not-applicable", "status": "trained-rejected", "trainable": false }, { "contamination_status": "exact task-id clear for canonical train/validation/monitor payloads; semantic benchmark-family contamination is unresolved unless the entry audit says otherwise", "dataset_id": "combined-v5-500", "excluded_from_hf": [], "files": { "data/manifest.json": { "sha256": "fbae544e8c4d358eb49513707982c50f9ec9349eb813c47a1f7aa76a55d45281", "size_bytes": 836 }, "data/sft/train.jsonl": { "sha256": "ec99a3faad13d568940b54d910c11199af65df4dd7cc19955f2c9ba693869d59", "size_bytes": 3620606 }, "evidence.tar.gz": { "sha256": "74e6874b9f33e50fe0c760f03cdddda532fe7d1fd8b7ae6f6845162114d4ff5a", "size_bytes": 17646 } }, "fixed26_task_id_overlap": 0, "kind": "glm47-aider-dataset-entry", "license_status": "unresolved-private-preservation-only", "lineage": "182 verified anchors + 318 reviewed new rows", "oracle_visibility": "internal review or verification receipts may be present; no hidden benchmark/runtime oracle", "role": "Reviewed SFT candidate", "rows": "500 rows", "schema_version": 1, "split_semantics": "not-applicable", "status": "candidate-audit-limited", "trainable": false }, { "contamination_status": "exact task-id clear for canonical train/validation/monitor payloads; semantic benchmark-family contamination is unresolved unless the entry audit says otherwise", "dataset_id": "sft-v3-complement-530", "excluded_from_hf": [], "files": { "data/manifest.json": { "sha256": "cbb79e9bcdab5738a0b6bcf4eea9c522e20358d6c8f77f03d03283f68ad2cfa5", "size_bytes": 1145 }, "data/sft/train.jsonl": { "sha256": "805aa59bbc936ee20687a293ef47d2fb9bcaee12419c6539ecd6180dfab02089", "size_bytes": 2952783 }, "evidence.tar.gz": { "sha256": "8830fe405d227a2514c3017eb0b8aeb8689b930cff45245865924c82d2f70cf9", "size_bytes": 15550 } }, "fixed26_task_id_overlap": 0, "kind": "glm47-aider-dataset-entry", "license_status": "unresolved-private-preservation-only", "lineage": "401 original anchors + 129 verified complementary rows", "oracle_visibility": "internal review or verification receipts may be present; no hidden benchmark/runtime oracle", "role": "SFT training corpus; strongest assisted fixed-26 parent", "rows": "530 packaged; 520 consumed", "schema_version": 1, "split_semantics": "train-only; fixed-26 remains external evaluation", "status": "historical-trained-promoted-parent", "trainable": true }, { "contamination_status": "fixed-26 exact task-ID overlap is zero; holistic-260 prompt 5-gram audit passed, while semantic similarity for the inherited 530 remains bounded by its parent evidence", "dataset_id": "sft-v4-holistic-790", "excluded_from_hf": [], "files": { "data/manifest.json": { "sha256": "daa813de6bc7339353742f382921de6e3a1ce63f85d9a95942b09b3276844c4d", "size_bytes": 1669 }, "data/sft/train.jsonl": { "sha256": "8c4af0ce677e5745f59e74c9f3a6ac11aae72ca11ee4fe837ddafcd1e1b10dcb", "size_bytes": 4095506 }, "evidence.tar.gz": { "sha256": "998f4fdc2f7ee1d6c71520dc122a66a694652b3ff10523bedf3ef228e68441ab", "size_bytes": 195183 } }, "fixed26_task_id_overlap": 0, "kind": "glm47-aider-dataset-entry", "license_status": "unresolved-private-preservation-only", "lineage": "Exact ordered union of SFT v3 complement-530 (530 rows) and holistic-260-v1 (260 independently replayed rows), with 129 legacy labels normalized to task_id under explicit lineage", "oracle_visibility": "internal review or verification receipts may be present; no hidden benchmark/runtime oracle", "role": "SFT v4 training corpus; v3 parent plus verified holistic curriculum", "rows": "790 packaged; 780 consumed by one full-batch epoch at global batch 20", "schema_version": 1, "split_semantics": "train-only; fixed-26 remains external evaluation", "status": "train-ready-not-yet-trained", "trainable": true }, { "contamination_status": "fixed-26 exact task-ID overlap is zero; 222 source-550 rows repeat normalized v4 prompt-answer messages and 40 target-novel rows remain review", "dataset_id": "sft-v5-experimental-1340", "files": { "data/README.md": { "sha256": "9cbffc34e89b4f5746ec5a01ede98511f99388cf8d1052941a4757b40a1b8a7f", "size_bytes": 427 }, "data/SHA256SUMS": { "sha256": "12a4c8a23654b5f9f9548ebc1e1cb69d50ef196a35cd43678c94db178468be30", "size_bytes": 780 }, "data/manifest.json": { "sha256": "0906e1abcec775ca52362679fe39d83af9ac1744c984faf1125f4de0c7b2e130", "size_bytes": 1901 }, "data/manifests/sources.json": { "sha256": "48210b79ba5183f43236198115f8aeceb5339778967b60b1b8c0f871d96a5538", "size_bytes": 577 }, "data/reports/composition.json": { "sha256": "50f95392f30ef83d6da2d34f14a59222f730342320f18834b989d8ede3fe6ee3", "size_bytes": 410 }, "data/reports/overlap_audit.json": { "sha256": "8293edb70a19c7b742eefa911793033cad23956a42c8e99c88720786bdab7e80", "size_bytes": 30336 }, "data/reports/token_audit.json": { "sha256": "6f0ebba0883bb088a47b231145ff1cf9d9d0ba5aef9e2e99564aad9f1e09b58c", "size_bytes": 406 }, "data/reports/training_plan.json": { "sha256": "08d7de83f1c7186a21285e905a5444c36a3e6d3aa1a42e50ee55b97069ca2c14", "size_bytes": 393 }, "data/review/queue.jsonl": { "sha256": "0218d301ab4ce0e4f78981b0e1c81ada57f4601de9961b75c1d1e355146d7d6b", "size_bytes": 4537 }, "data/sft/train.jsonl": { "sha256": "a01a07c9d4e2706683814a3d5afc2bcd47172ff92b08f15e2c673729f185bd66", "size_bytes": 6836578 } }, "fixed26_task_id_overlap": 0, "kind": "glm47-aider-dataset-entry", "license_status": "unresolved-private-preservation-only", "lineage": "Exact ordered union: SFT v4 holistic 790 plus source-550 v1; all overlaps and review rows are disclosed in package reports", "oracle_visibility": "internal review receipts are present; no hidden fixed-26 benchmark oracle", "role": "Experimental SFT v5 corpus: ordered union of v4-790 and source-550", "rows": "1340 packaged and consumed per epoch at global batch 20", "schema_version": 1, "split_semantics": "train-only; fixed-26 remains external evaluation", "status": "experimental-train-authorized-not-quality-promoted", "trainable": true }, { "contamination_status": "fixed-26 exact task-ID overlap is zero; exact normalized prompt, answer, task-ID, and message duplicates are zero", "dataset_id": "sft-v6-audited-2000", "files": { "data/README.md": { "sha256": "ef34de87ec6e08427f8b75b7535cd13b7542a53fae0cd02487b723f10fd8027c", "size_bytes": 932 }, "data/SHA256SUMS": { "sha256": "0051983019ca3c3c676e565384ac565cd06d7d547429328801664248869ac35b", "size_bytes": 2413 }, "data/inputs/candidate_receipts.jsonl": { "sha256": "14bc6a23f1cef06a166c90d7c652e868b494c64ada9865f40729415b3fa70631", "size_bytes": 2122948 }, "data/inputs/candidate_variants.jsonl": { "sha256": "71d20a8175919af8d12d715796accfc57db8d5eea1c1e10c199d24c179d7ad29", "size_bytes": 2828836 }, "data/inputs/fixed26_manifest.json": { "sha256": "4c2a35dc8841bc8c5d86fdd4f04539ff917d105efacc7ea46117381b9e6056b7", "size_bytes": 5379 }, "data/inputs/holistic_tests.jsonl": { "sha256": "453eabe8ac7968298b6cb1e321203966bc760aa9e138f8463499af6d636e2099", "size_bytes": 221941 }, "data/inputs/holistic_token_audit.json": { "sha256": "c56446a83e5e6c3cf082cad0890072ad3b3cb9b5d33d77b6d7edd560f6e03fe7", "size_bytes": 30915 }, "data/inputs/holistic_train.jsonl": { "sha256": "03d42d3c04dea452bb9d094366bd504a3da8ffa01cf04a66e220a15a1b93b03a", "size_bytes": 1143771 }, "data/inputs/holistic_verification.jsonl": { "sha256": "f9724d1452c0d1ebaa8b59a571131726b033a190640b1c4bae869cd4d9d5571b", "size_bytes": 199674 }, "data/inputs/pass1_rows.jsonl": { "sha256": "f4addfdbe8652714e46b7c2a7123f116b449577eaee3676f74c272eeef90a904", "size_bytes": 226223 }, "data/inputs/pass1_verification.jsonl": { "sha256": "a36c9cd9c327230749bb1d87186b2cd439ff37346d72212fabe6da9f5d3d3041", "size_bytes": 26150 }, "data/inputs/v5_overlap_audit.json": { "sha256": "8293edb70a19c7b742eefa911793033cad23956a42c8e99c88720786bdab7e80", "size_bytes": 30336 }, "data/inputs/v5_review_queue.jsonl": { "sha256": "0218d301ab4ce0e4f78981b0e1c81ada57f4601de9961b75c1d1e355146d7d6b", "size_bytes": 4537 }, "data/inputs/v5_train.jsonl": { "sha256": "a01a07c9d4e2706683814a3d5afc2bcd47172ff92b08f15e2c673729f185bd66", "size_bytes": 6836578 }, "data/manifest.json": { "sha256": "04b51b576e34ba90355cee0268297a0ee9093da239fd98db346dd023ef8d8c32", "size_bytes": 3472 }, "data/manifests/selected_train.jsonl": { "sha256": "f108e59d05b6f36ae66cd0c186fb41bc911ef83f346f920aeb1fb2a8203faf3c", "size_bytes": 338614 }, "data/manifests/sources.json": { "sha256": "056f7c6583aefdea769aad48aa1ceaa190cd9beeb9c0e2ab5ed068fdc0505d24", "size_bytes": 2075 }, "data/receipts/composition_verification.jsonl": { "sha256": "13adec22899656acfe9d6a40ed76d2d4c18d5e08809a29f25b11b70bf6a40cca", "size_bytes": 746943 }, "data/rejected/composition_pairs.jsonl": { "sha256": "b1ce92d8319a027b59467ab73975a17a7af8351c0fa2822e52609073380d0fac", "size_bytes": 1861 }, "data/rejected/rows.jsonl": { "sha256": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855", "size_bytes": 0 }, "data/reports/composition.json": { "sha256": "a0eb3c58025ecf5d1f4c8daed5bbc35c428aa312cb380e46ce2da4a75ad71a68", "size_bytes": 12070 }, "data/reports/contamination_audit.json": { "sha256": "812e9d78a33479cab488cf9571c9a79b1fbf458d621f9df941df02876401806e", "size_bytes": 284 }, "data/reports/duplicate_audit.json": { "sha256": "926be21881cee68142050f81d54f83a93606aec620ec3cc659f60b608b7f0758", "size_bytes": 510 }, "data/reports/row_catalog.jsonl": { "sha256": "be80c71a1da914e064686b9e714bfa51cb6a6e6c23131d075473c582653f5802", "size_bytes": 1181238 }, "data/reports/token_audit.json": { "sha256": "2e8257bb66ec6f1b6ba1a9422256f24a97468630006f5959d9fdbef6476ef546", "size_bytes": 274409 }, "data/review/queue.jsonl": { "sha256": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855", "size_bytes": 0 }, "data/sft/train.jsonl": { "sha256": "debe8081a7f780afa23942e7a0ab06358aeaaba6f6006557eb5cff93779d9995", "size_bytes": 14939942 } }, "fixed26_task_id_overlap": 0, "kind": "glm47-aider-dataset-entry", "license_status": "private-research-corpus", "lineage": "949 quality-cleared target-distinct v5 rows, 70 executable pass1 rows, 3 unused mutation-adequate direct rows, and 978 current-replay cross-family multi-file compositions", "oracle_visibility": "component execution receipts are included; no fixed-26 answer, test, rubric, or grader state is included", "role": "SFT v6 corpus for single-turn C++17 repository editing in Aider whole-file format", "rows": "2000 packaged and consumed per epoch at global batch 20", "schema_version": 1, "split_semantics": "train-only; fixed-26 remains external evaluation", "status": "trained-rejected", "train_sha256": "debe8081a7f780afa23942e7a0ab06358aeaaba6f6006557eb5cff93779d9995", "trainable": false }, { "contamination_status": "fixed-26 contamination gate passed; see data/manifest.json", "dataset_id": "sft-v7-corrected-2000", "files": { "data/README.md": { "sha256": "392600a4804af39b3e2e5166958dbf0dd6ff1f71b1a17a6f31b42e52f6bbf4b1", "size_bytes": 1868 }, "data/SHA256SUMS": { "sha256": "e758f700101505f209c78d3c7265a802cb12a6fdf6b9eb2b868428e2062467ec", "size_bytes": 3526 }, "data/inputs/candidate_receipts.jsonl": { "sha256": "14bc6a23f1cef06a166c90d7c652e868b494c64ada9865f40729415b3fa70631", "size_bytes": 2122948 }, "data/inputs/candidate_variants.jsonl": { "sha256": "71d20a8175919af8d12d715796accfc57db8d5eea1c1e10c199d24c179d7ad29", "size_bytes": 2828836 }, "data/inputs/codecontests_candidates.jsonl": { "sha256": "7cf198aaaa94af6786909299da8612521efd0f06048d189ec5e0f6def2c6c760", "size_bytes": 53831211 }, "data/inputs/codecontests_source_manifest.json": { "sha256": "d5355d54a0323fd9fb66fc7c9d539e0c6e64e65241e9fcede6503accdd0d4813", "size_bytes": 1778 }, "data/inputs/codecontests_verification.jsonl": { "sha256": "38f19487e137d87768b14b381ad45bc0806471cae871437503e6e576ffb4819c", "size_bytes": 1133417 }, "data/inputs/fixed26_manifest.json": { "sha256": "4c2a35dc8841bc8c5d86fdd4f04539ff917d105efacc7ea46117381b9e6056b7", "size_bytes": 5379 }, "data/inputs/fixed26_signatures.json": { "sha256": "59de33afbc023d625a41b5d77ebda43896b9c2cd5bb3b5ea4959f9202739771f", "size_bytes": 448639 }, "data/inputs/pass1_rows.jsonl": { "sha256": "f4addfdbe8652714e46b7c2a7123f116b449577eaee3676f74c272eeef90a904", "size_bytes": 226223 }, "data/inputs/pass1_verification.jsonl": { "sha256": "a36c9cd9c327230749bb1d87186b2cd439ff37346d72212fabe6da9f5d3d3041", "size_bytes": 26150 }, "data/inputs/v5_overlap_audit.json": { "sha256": "8293edb70a19c7b742eefa911793033cad23956a42c8e99c88720786bdab7e80", "size_bytes": 30336 }, "data/inputs/v5_review_queue.jsonl": { "sha256": "0218d301ab4ce0e4f78981b0e1c81ada57f4601de9961b75c1d1e355146d7d6b", "size_bytes": 4537 }, "data/inputs/v5_train.jsonl": { "sha256": "a01a07c9d4e2706683814a3d5afc2bcd47172ff92b08f15e2c673729f185bd66", "size_bytes": 6836578 }, "data/inputs/v6_composition_verification.jsonl": { "sha256": "13adec22899656acfe9d6a40ed76d2d4c18d5e08809a29f25b11b70bf6a40cca", "size_bytes": 746943 }, "data/inputs/v6_train.jsonl": { "sha256": "debe8081a7f780afa23942e7a0ab06358aeaaba6f6006557eb5cff93779d9995", "size_bytes": 14939942 }, "data/manifest.json": { "sha256": "98912424174e35bfa3961e45ba55282e0cd571e0c19c0ce42f8148ba638451e5", "size_bytes": 9352 }, "data/manifests/selected_train.jsonl": { "sha256": "481c074859cf9636fbaddf95b4d3f0551155a07d446464c744cd4279326749a3", "size_bytes": 2123239 }, "data/manifests/sources.json": { "sha256": "baaf4559bf98aed2eff90e9b542b7c32764b0c8674fbcdb215cab4bb5e2a593c", "size_bytes": 2495 }, "data/manifests/verification_receipt_index.jsonl": { "sha256": "39f2661ad696513ee0dd4c2963303b306ed2fae05448d085919ab54f9eb1039b", "size_bytes": 440960 }, "data/receipts/codecontests_verification.jsonl": { "sha256": "fb754e64750e7e7c93ac54e4551542bcbbddf2ee23c8c96cba34e5ca20aa46d8", "size_bytes": 1019914 }, "data/receipts/composition_verification.jsonl": { "sha256": "72462ee163e7d234608cc040e17b994375c7ce4038d32186569ffd5ae593e8b0", "size_bytes": 190886 }, "data/receipts/deterministic_rebuild.json": { "sha256": "97a5280e036352a134bd3835bd16133e6559ee5188b0537cddd3adeb47141578", "size_bytes": 573 }, "data/receipts/inherited_anchor_admission.jsonl": { "sha256": "594b826f5482d80bfd6867044cf1ad4242a5a4955c66e5584bff57af33b5f367", "size_bytes": 356073 }, "data/rejected/codecontests_candidates.jsonl": { "sha256": "4f64070663e3686e5a851e21ffe43a07f16a620ee4c0bb2e126a90dbd6b51d2d", "size_bytes": 17539 }, "data/reports/aider_format_audit.json": { "sha256": "819eaa3c71a37a71a3d70687e74d404b084e1b5bb0c2a28df3152bd9b4dd3932", "size_bytes": 289 }, "data/reports/block_reuse_audit.json": { "sha256": "45d4a3de7a120c747205b89a1da9bf8836b7d47d494cd106dc09ea53e3f32b58", "size_bytes": 160750 }, "data/reports/codecontests_contamination_prefilter.json": { "sha256": "7929cb91ae182c53a605c05661d3401b37bd5caa428a7635270ebab8aaa3aa07", "size_bytes": 2116 }, "data/reports/contamination_audit.json": { "sha256": "5c50c606af63c0e6d47f2b7996a552127cca672c09b971154704bbb82e66550b", "size_bytes": 1213 }, "data/reports/duplicate_audit.json": { "sha256": "d8cd7c67bf71f182e23b58f9ea6606883e309af53bbf9e013b322b290d7b785c", "size_bytes": 618 }, "data/reports/family_mixture_audit.json": { "sha256": "41a3e22c3d8591af06bac73259adb42a1c043f53a048ec3af2e9e67d662f81cd", "size_bytes": 70377 }, "data/reports/row_catalog.jsonl": { "sha256": "481c074859cf9636fbaddf95b4d3f0551155a07d446464c744cd4279326749a3", "size_bytes": 2123239 }, "data/reports/train_admission_audit.json": { "sha256": "6bfefe74b7b32da311816df46b6f1caa46e0f9d666da6aae05e1958d4e0d4e1c", "size_bytes": 1913138 }, "data/reports/v5_disposition_ledger.jsonl": { "sha256": "8e9ece5d306c2f3a109e3e7bd66f716939589b92b0927c8a1b21bc7a9af72885", "size_bytes": 702228 }, "data/reports/verification_receipt_reference_audit.json": { "sha256": "456850bad234d8def849c30a262aa02e9e91c36dd61e62e28183db219b92c3af", "size_bytes": 328 }, "data/review/queue.jsonl": { "sha256": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855", "size_bytes": 0 }, "data/sft/train.jsonl": { "sha256": "2bdf04491c776a74ecffc3a896dec3b9d9783a693dbc3981868b2153154c59bf", "size_bytes": 12166603 } }, "fixed26_task_id_overlap": 0, "kind": "glm47-aider-dataset-entry", "license_status": "mixed-source-review-required", "lineage": "Corrected v7 lineage is recorded in data/manifest.json", "oracle_visibility": "row verification receipts are included; no fixed-26 answer, test, rubric, or grader state is included", "role": "Corrected SFT v7 corpus for single-turn C++17 repository editing in Aider whole-file format", "rows": "2000 admitted train rows", "schema_version": 1, "source_manifest_sha256": "98912424174e35bfa3961e45ba55282e0cd571e0c19c0ce42f8148ba638451e5", "split_semantics": "train-only; fixed-26 remains external evaluation", "status": "row-level-train-ready-model-benefit-not-yet-evaluated", "train_sha256": "2bdf04491c776a74ecffc3a896dec3b9d9783a693dbc3981868b2153154c59bf", "trainable": true }, { "contamination_status": "exact task-id clear for canonical train/validation/monitor payloads; semantic benchmark-family contamination is unresolved unless the entry audit says otherwise", "dataset_id": "pass1-skills-600", "excluded_from_hf": [], "files": { "data/manifest.json": { "sha256": "d626da77e218720cec1ac078a37a98cbd037de3aed723bd4aca85cec45a10c51", "size_bytes": 2121 }, "data/sft/train.jsonl": { "sha256": "6bf9a449623c1f3a9087155d0ce229de8afb0842a6ea1695703394775feb4aac", "size_bytes": 3179006 }, "evidence.tar.gz": { "sha256": "594e62a5666f3bebaf6fa672d69ed003cd4f9bdffe8a96ab9132f8db41333056", "size_bytes": 48473 } }, "fixed26_task_id_overlap": 0, "kind": "glm47-aider-dataset-entry", "license_status": "unresolved-private-preservation-only", "lineage": "SFT v3 complement-530 + 70 synthetic skill rows", "oracle_visibility": "internal review or verification receipts may be present; no hidden benchmark/runtime oracle", "role": "Proposed direct-success SFT successor", "rows": "600 rows", "schema_version": 1, "split_semantics": "not-applicable", "status": "evaluated-candidate", "trainable": false }, { "contamination_status": "exact task-id clear for canonical train/validation/monitor payloads; semantic benchmark-family contamination is unresolved unless the entry audit says otherwise", "dataset_id": "signal-v2", "excluded_from_hf": [ "private/ reference rows", "staging/ generated tests and fixtures" ], "files": { "data/manifest.json": { "sha256": "a4be48b43019ee7fad70ea4ae865307e886cc1821a5337fa86b6691a6fd6f7ce", "size_bytes": 1255 }, "data/sft/train.jsonl": { "sha256": "61daff8a833955084f882782b70448739d155135a15057c765909bb963c4ea30", "size_bytes": 1556938 }, "data/validation/validation.jsonl": { "sha256": "14dd2e4c498fc5e06cd41d2d285797a0a38dff93fc9cb240950972f2661a9f73", "size_bytes": 52471 }, "evidence.tar.gz": { "sha256": "e2ba27f1a06e025da14a92c574637aefffd75ced31b488cffcd515e0f29763e6", "size_bytes": 248503 } }, "fixed26_task_id_overlap": 0, "kind": "glm47-aider-dataset-entry", "license_status": "unresolved-private-preservation-only", "lineage": "Answer-blind verified base tasks plus authored trajectories", "oracle_visibility": "internal review or verification receipts may be present; no hidden benchmark/runtime oracle", "role": "Signal-focused SFT candidate", "rows": "150 train + 4 validation", "schema_version": 1, "split_semantics": "not-applicable", "status": "candidate", "trainable": false }, { "contamination_status": "exact task-id clear for canonical train/validation/monitor payloads; semantic benchmark-family contamination is unresolved unless the entry audit says otherwise", "dataset_id": "signal-v3", "excluded_from_hf": [ "private/ reference rows", "staging/ generated tests and fixtures" ], "files": { "data/manifest.json": { "sha256": "8a5b9ab29aa740ab642f1213ac89df94b4ffcfe373e3ba332b0eff554e35cf3e", "size_bytes": 1437 }, "data/sft/train.jsonl": { "sha256": "11d75ca74088f3dda3f8ba42abef7cfe98b4a59375aeadc652886ec7094b184f", "size_bytes": 2766459 }, "data/validation/validation.jsonl": { "sha256": "f3d5cf7f67ae3c9d874ab7be72680b2ccad4038bf55bc8bdffbbb0fe8ac6b5ce", "size_bytes": 83330 }, "evidence.tar.gz": { "sha256": "3da47ae3b87948b433ed79d3de653676f176a68196e83e7c70e5dc7ecd77c152", "size_bytes": 367721 } }, "fixed26_task_id_overlap": 0, "kind": "glm47-aider-dataset-entry", "license_status": "unresolved-private-preservation-only", "lineage": "Answer-blind verified base tasks plus authored trajectories", "oracle_visibility": "internal review or verification receipts may be present; no hidden benchmark/runtime oracle", "role": "Expanded signal-focused SFT corpus", "rows": "249 train + 6 validation", "schema_version": 1, "split_semantics": "not-applicable", "status": "trained-evaluated", "trainable": false }, { "contamination_status": "exact task-id clear for canonical train/validation/monitor payloads; semantic benchmark-family contamination is unresolved unless the entry audit says otherwise", "dataset_id": "reverify-715", "excluded_from_hf": [], "files": { "data/manifest.json": { "sha256": "d2762d4fe1642f19816d79a1a448469e39f75583e2bfa0e129cf3fe354d0a41b", "size_bytes": 284588 }, "data/sft/train.jsonl": { "sha256": "741d075a85f1f2f9c76b1ee03ea21e6355f9bb94e352563411d744be91c2bf9c", "size_bytes": 3884675 }, "data/source/aider-tasks-reverify-sft.zip": { "sha256": "f90ff7c6c595b91f06b42c89e8d224f7f0d3786e5376ff17384aab59a13e9d38", "size_bytes": 478473 } }, "fixed26_task_id_overlap": 0, "kind": "glm47-aider-dataset-entry", "license_status": "unresolved-private-preservation-only", "lineage": "Reverified Aider task projection", "oracle_visibility": "internal review or verification receipts may be present; no hidden benchmark/runtime oracle", "role": "Audit projection and raw-concat source", "rows": "715 rows", "schema_version": 1, "split_semantics": "not-applicable", "status": "audit-only", "trainable": false }, { "contamination_status": "not applicable to a canonical train/validation/monitor payload; row-level audit decisions are preserved as evidence", "dataset_id": "reverify-audit", "excluded_from_hf": [ "evidence/ benchmark trees and transcripts" ], "files": { "evidence.tar.gz": { "sha256": "d708818a2c058b84e38b65471098df49ae31ee4203a430024002e6c44713f5e3", "size_bytes": 162195 } }, "fixed26_task_id_overlap": null, "kind": "glm47-aider-dataset-entry", "license_status": "unresolved-private-preservation-only", "lineage": "Lineage evidence for reverify and SFT v2 remediation", "oracle_visibility": "internal review or verification receipts may be present; no hidden benchmark/runtime oracle", "role": "Dataset quality/audit ledger", "rows": "row-level catalog and decisions", "schema_version": 1, "split_semantics": "not-applicable", "status": "audit-only", "trainable": false }, { "contamination_status": "not applicable to a canonical train/validation/monitor payload; row-level audit decisions are preserved as evidence", "dataset_id": "regression-audit", "excluded_from_hf": [ "evidence/ benchmark trees and transcripts" ], "files": { "evidence.tar.gz": { "sha256": "90e2ea3af5ed0304a3c330f4f2a7c63616a0470c7b692412223668177767303a", "size_bytes": 393224 } }, "fixed26_task_id_overlap": null, "kind": "glm47-aider-dataset-entry", "license_status": "unresolved-private-preservation-only", "lineage": "Lineage evidence for reverify and SFT v2 remediation", "oracle_visibility": "internal review or verification receipts may be present; no hidden benchmark/runtime oracle", "role": "Dataset quality/audit ledger", "rows": "row-level catalog and decisions", "schema_version": 1, "split_semantics": "not-applicable", "status": "audit-only", "trainable": false }, { "contamination_status": "unresolved candidate-generation workbench; not approved as a canonical training payload", "dataset_id": "combined-v5-build-workbench", "excluded_from_hf": [ "executable_test_receipts.jsonl", "judge and mutation oracle outputs", "semantic_judgments.jsonl", "semantic_verifier/ blind packets and tests", "supplemental_blind_tests_v1/", "test_staging.jsonl" ], "files": { "data/candidate_pool.jsonl": { "sha256": "b62c13da6f34ba6f13c579c12a6693f3ab433681716c8bf5ed6a15524a2be3ee", "size_bytes": 238299 }, "data/candidate_variants.jsonl": { "sha256": "71d20a8175919af8d12d715796accfc57db8d5eea1c1e10c199d24c179d7ad29", "size_bytes": 2828836 }, "evidence.tar.gz": { "sha256": "844892b8b33e449efff01c7581cad261a7964db102285ed60afd2e44010c39f4", "size_bytes": 11537 } }, "fixed26_task_id_overlap": null, "kind": "glm47-aider-dataset-entry", "license_status": "unresolved-private-preservation-only", "lineage": "Intermediate materialization behind combined-v5-500", "oracle_visibility": "internal review or verification receipts may be present; no hidden benchmark/runtime oracle", "role": "Candidate-generation workbench", "rows": "candidate pool and candidate variants", "schema_version": 1, "split_semantics": "not-applicable", "status": "lineage-only-not-a-training-dataset", "trainable": false }, { "contamination_status": "exact task-id clear for canonical train/validation/monitor payloads; semantic benchmark-family contamination is unresolved unless the entry audit says otherwise", "dataset_id": "rl-full-replay-253", "excluded_from_hf": [ "rl_tasks/ task fixtures and grader tests" ], "files": { "data/eval/train_monitor.jsonl": { "sha256": "fe5e26207c5da5578e24fa9ab772aa170789db4d57e3749602b810fc6020df7f", "size_bytes": 206054 }, "data/grpo/train.jsonl": { "sha256": "1ecb721e0534a587ca00420b1e124929e72b61f669963ff4cfbde18ae0617e8a", "size_bytes": 1608324 }, "data/manifest.json": { "sha256": "7bcc98a1cbd4d2cbf1f8c9b7e3e1ff3ffff8ed84411b2d70824810664cfafbe7", "size_bytes": 20506 } }, "fixed26_task_id_overlap": 0, "kind": "glm47-aider-dataset-entry", "license_status": "unresolved-private-preservation-only", "lineage": "Projection of gated answer-free Aider-style C++ RL corpus", "oracle_visibility": "internal review or verification receipts may be present; no hidden benchmark/runtime oracle", "role": "GRPO prompt and monitor projection", "rows": "253 train + 32 monitor", "schema_version": 1, "split_semantics": "not-applicable", "status": "prepared-not-promoted", "trainable": false }, { "contamination_status": "not applicable to a canonical training payload; this entry contains only the difficulty-selection audit", "dataset_id": "rl-difficulty-20260722", "excluded_from_hf": [], "files": { "data/difficulty_rows.json": { "sha256": "be17d6be3448529661f565dce5a9048e9c025bf977c0fc9eaa0d076f224ee2cc", "size_bytes": 80261 }, "data/report.json": { "sha256": "73dd3946e6860f09faf22063409df42d478c9194d1bf9c5858658984020110b6", "size_bytes": 9164 }, "evidence.tar.gz": { "sha256": "bcca48a8e3be6efc57ff4de504062ae91254a8780f81b4d27291a6bee000d04e", "size_bytes": 673 } }, "fixed26_task_id_overlap": null, "kind": "glm47-aider-dataset-entry", "license_status": "unresolved-private-preservation-only", "lineage": "Selection audit that produced the repaired RL v2 subset", "oracle_visibility": "internal review or verification receipts may be present; no hidden benchmark/runtime oracle", "role": "Difficulty filter and selection audit", "rows": "253 assessed; 169 kept; 84 dropped", "schema_version": 1, "split_semantics": "not-applicable", "status": "audit-only-filter", "trainable": false }, { "contamination_status": "exact task-id clear for canonical train/validation/monitor payloads; semantic benchmark-family contamination is unresolved unless the entry audit says otherwise", "dataset_id": "rl-v2-169", "excluded_from_hf": [ "rl_tasks/ task fixtures and grader tests" ], "files": { "data/eval/train_monitor.jsonl": { "sha256": "4553c162306412b069653af28ed13b38fad5fb035db0a1f3fe8396dacbf6a25a", "size_bytes": 135300 }, "data/grpo/train.jsonl": { "sha256": "bb7472bb551d95e180d391351567372c62fd72bee33a0ca4cf8186d294f2c882", "size_bytes": 1070405 }, "data/manifest.json": { "sha256": "65b50a6532abc87f78cf43e673b07625c4056a62ecfdacc9a49f5849e6318105", "size_bytes": 14722 } }, "fixed26_task_id_overlap": 0, "kind": "glm47-aider-dataset-entry", "license_status": "unresolved-private-preservation-only", "lineage": "Difficulty-filtered subset of 253-task Aider-style C++ RL corpus", "oracle_visibility": "internal review or verification receipts may be present; no hidden benchmark/runtime oracle", "role": "Actual repaired two-epoch GRPO input", "rows": "169 train + 22 monitor", "schema_version": 1, "split_semantics": "monitor IDs overlap training tasks; monitor is non-gradient training telemetry, not a held-out validation split", "status": "historical-trained", "trainable": true }, { "contamination_status": "exact task-id clear for canonical train/validation/monitor payloads; semantic benchmark-family contamination is unresolved unless the entry audit says otherwise", "dataset_id": "rl-full-253-july21", "excluded_from_hf": [ "rl_tasks/ task fixtures and grader tests" ], "files": { "data/eval/train_monitor.jsonl": { "sha256": "c4e1471a53a609fa1c310d31c139693ffc21424b6a12891884fbd20844690492", "size_bytes": 111993 }, "data/grpo/train.jsonl": { "sha256": "6b293a6bd8925eef5497a4c198e7c207ee74e867ba75f92dc3e41faab592f7c5", "size_bytes": 917306 }, "data/manifest.json": { "sha256": "5dc240d6658097f4014ddb841e1381dc0685fd39cdd83a6b13054057d9d5d620", "size_bytes": 20454 } }, "fixed26_task_id_overlap": 0, "kind": "glm47-aider-dataset-entry", "license_status": "unresolved-private-preservation-only", "lineage": "Pre-filter full Aider-style C++ RL projection", "oracle_visibility": "internal review or verification receipts may be present; no hidden benchmark/runtime oracle", "role": "Earlier GRPO input preserved with the July 21 run", "rows": "253 train + 32 monitor", "schema_version": 1, "split_semantics": "monitor IDs overlap training tasks; monitor is non-gradient training telemetry, not a held-out validation split", "status": "historical-trained-not-promoted", "trainable": false } ], "existing_gated_repositories": [ { "dataset_id": "pie-cpp-posttraining", "repo": "TokenBender/glm47-pie-cpp-posttraining-data", "revision": "09bc0276a0ff8ab84a8db81880ca7f739057e654", "status": "existing-gated" }, { "boundary": "Contains rubrics and hidden executable tests; approve only RL runtime/service identities, not ordinary corpus consumers.", "dataset_id": "aider-cpp-rl-runtime-v2", "repo": "TokenBender/glm47-aider-cpp-rl-tasks", "revision": "d8f86f752685d5ddc6cece2a08ea8851b395ee83", "status": "existing-gated-runtime-oracle-restricted" }, { "dataset_id": "rl8-validity-6-plus-2", "repo": "TokenBender/glm47-aider-rl8-validity-rollouts-20260723", "revision": "11451eeec5261eecdbca75e95ce190a632783654", "status": "existing-gated-framework-validity-only" }, { "dataset_id": "rl-v3-failed-rollouts", "repo": "TokenBender/glm47-aider-rl-v3-rollouts-20260722", "revision": "bcd3b747c3d5881921f269a8690def5986a56bed", "status": "existing-gated-failed-infrastructure" } ], "kind": "glm47-aider-posttraining-data-catalog", "known_missing_payloads": [ { "dataset_id": "sft-v1-321", "impact": "These projections were not consumed by the recorded SFT run; the exact 321-row train file and 320-row consumption order are preserved.", "payloads": [ "40-row validation projection", "40-row test projection", "split_manifest.json" ] } ], "policy": { "all_extant_project_payloads_mapped": true, "gating_is_not_redistribution_authority": true, "hidden_tests_oracles_republished": false, "repo_access": "private-manual-gated" }, "schema_version": 1 }