{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2024:UJXXSEMUFSV3FLT7TK7VRHMFUC","merge_version":"pith-open-graph-merge-v1","event_count":2,"valid_event_count":2,"invalid_event_count":0,"equivocation_count":0,"current":{"canonical_record":{"metadata":{"abstract_canon_sha256":"335b86f497a45c80eb211d711a640b7ed39d714988b1318ba0d700fed18ed3fa","cross_cats_sorted":["cs.LG"],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2024-07-23T16:13:22Z","title_canon_sha256":"58f68631091b228389dc12bff3e5788dbd7fc0ba7841838d0f9fc7e6bff20354"},"schema_version":"1.0","source":{"id":"2407.16607","kind":"arxiv","version":4}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2407.16607","created_at":"2026-07-05T09:42:28Z"},{"alias_kind":"arxiv_version","alias_value":"2407.16607v4","created_at":"2026-07-05T09:42:28Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2407.16607","created_at":"2026-07-05T09:42:28Z"},{"alias_kind":"pith_short_12","alias_value":"UJXXSEMUFSV3","created_at":"2026-07-05T09:42:28Z"},{"alias_kind":"pith_short_16","alias_value":"UJXXSEMUFSV3FLT7","created_at":"2026-07-05T09:42:28Z"},{"alias_kind":"pith_short_8","alias_value":"UJXXSEMU","created_at":"2026-07-05T09:42:28Z"}],"graph_snapshots":[{"event_id":"sha256:5ac7a8f184b805f79af4048634d0012241f3aa314b9162fdb3b81b23030d0c61","target":"graph","created_at":"2026-07-05T09:42:28Z","signer":{"key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signer_id":"pith.science","signer_type":"pith_registry"},"payload":{"graph_snapshot":{"author_claims":{"count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57","strong_count":0},"builder_version":"pith-number-builder-2026-05-17-v1","claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"formal_canon":{"evidence_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"integrity":{"available":true,"clean":true,"detectors_run":[],"endpoint":"/pith/2407.16607/integrity.json","findings":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"The pretraining data of today's strongest language models is opaque; in particular, little is known about the proportions of various domains or languages represented. In this work, we tackle a task which we call data mixture inference, which aims to uncover the distributional make-up of training data. We introduce a novel attack based on a previously overlooked source of information: byte-pair encoding (BPE) tokenizers, used by the vast majority of modern language models. Our key insight is that the ordered list of merge rules learned by a BPE tokenizer naturally reveals information about the ","authors_text":"Alisa Liu, Jonathan Hayase, Noah A. Smith, Sewoong Oh, Yejin Choi","cross_cats":["cs.LG"],"headline":"","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2024-07-23T16:13:22Z","title":"Data Mixture Inference: What do BPE Tokenizers Reveal about their Training Data?"},"references":{"count":0,"internal_anchors":0,"resolved_work":0,"sample":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2407.16607","kind":"arxiv","version":4},"verdict":{"created_at":null,"id":null,"model_set":{},"one_line_summary":"","pipeline_version":null,"pith_extraction_headline":"","strongest_claim":"","weakest_assumption":""}},"verdict_id":null}}],"author_attestations":[],"timestamp_anchors":[],"storage_attestations":[],"citation_signatures":[],"replication_records":[],"corrections":[],"mirror_hints":[],"record_created":{"event_id":"sha256:aa7fc7c75c1e4d5d15677575712702bc6c94b4429277ab61a521ab93a71e46c9","target":"record","created_at":"2026-07-05T09:42:28Z","signer":{"key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signer_id":"pith.science","signer_type":"pith_registry"},"payload":{"attestation_state":"computed","canonical_record":{"metadata":{"abstract_canon_sha256":"335b86f497a45c80eb211d711a640b7ed39d714988b1318ba0d700fed18ed3fa","cross_cats_sorted":["cs.LG"],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2024-07-23T16:13:22Z","title_canon_sha256":"58f68631091b228389dc12bff3e5788dbd7fc0ba7841838d0f9fc7e6bff20354"},"schema_version":"1.0","source":{"id":"2407.16607","kind":"arxiv","version":4}},"canonical_sha256":"a26f7911942cabb2ae7f9abf589d85a0becc0976eda5dbeede26dec28551f65e","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"a26f7911942cabb2ae7f9abf589d85a0becc0976eda5dbeede26dec28551f65e","first_computed_at":"2026-07-05T09:42:28.239003Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-07-05T09:42:28.239003Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"aBZ6aeNSwZ6yT9N57IzlNvux7WCU2ff6/GQUti1t0+GdZkK3O2A3tSihP7RTJOY7Vad2GUkLOtQpTeGROVd1Cg==","signature_status":"signed_v1","signed_at":"2026-07-05T09:42:28.239498Z","signed_message":"canonical_sha256_bytes"},"source_id":"2407.16607","source_kind":"arxiv","source_version":4}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:aa7fc7c75c1e4d5d15677575712702bc6c94b4429277ab61a521ab93a71e46c9","sha256:5ac7a8f184b805f79af4048634d0012241f3aa314b9162fdb3b81b23030d0c61"],"state_sha256":"6690bd5b2690af45a7cb8bb2349995b42f22f3db9c0056eab96cd48d566ec91f"}