{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2026:7KV2TX4SGGWHL42GIHWMI5PL3P","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":"a4eb8756ec90079c47ff7deb3085703cf3dc7bc4f068cec585aad1704d97d8c9","cross_cats_sorted":["cs.LG"],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2026-02-04T07:36:46Z","title_canon_sha256":"e920092daf9b8b6cd485968f985685df05792529c32cedad61380bddef078911"},"schema_version":"1.0","source":{"id":"2602.04289","kind":"arxiv","version":2}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2602.04289","created_at":"2026-05-17T23:39:00Z"},{"alias_kind":"arxiv_version","alias_value":"2602.04289v2","created_at":"2026-05-17T23:39:00Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2602.04289","created_at":"2026-05-17T23:39:00Z"},{"alias_kind":"pith_short_12","alias_value":"7KV2TX4SGGWH","created_at":"2026-05-18T12:33:37Z"},{"alias_kind":"pith_short_16","alias_value":"7KV2TX4SGGWHL42G","created_at":"2026-05-18T12:33:37Z"},{"alias_kind":"pith_short_8","alias_value":"7KV2TX4S","created_at":"2026-05-18T12:33:37Z"}],"graph_snapshots":[{"event_id":"sha256:27206fbfa882b7306a4b048806e08e55347875247116ff7a813df833a112f45b","target":"graph","created_at":"2026-05-17T23:39:00Z","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":4,"items":[{"attestation":"unclaimed","claim_id":"C1","kind":"strongest_claim","source":"verdict.strongest_claim","status":"machine_extracted","text":"proxy compression substantially improves training efficiency and significantly outperforms pure byte-level baselines given fixed compute budgets. As model scale increases, these gains become more pronounced, and proxy-trained models eventually match or surpass tokenizer approaches, all while operating solely on raw bytes."},{"attestation":"unclaimed","claim_id":"C2","kind":"weakest_assumption","source":"verdict.weakest_assumption","status":"machine_extracted","text":"that the internal alignment learned during joint training on compressed and raw views transfers effectively to pure raw-byte inference without performance degradation or the need for continued compressed inputs."},{"attestation":"unclaimed","claim_id":"C3","kind":"one_line_summary","source":"verdict.one_line_summary","status":"machine_extracted","text":"Proxy compression trains language models on both raw bytes and compressed sequences to enable efficient training with raw-byte inference at test time."},{"attestation":"unclaimed","claim_id":"C4","kind":"headline","source":"verdict.pith_extraction.headline","status":"machine_extracted","text":"Proxy compression trains language models jointly on raw bytes and compressed sequences so they can use efficient inputs during training yet run purely on raw bytes at inference."}],"snapshot_sha256":"bd42ab1bbd133da7cccab68460c3a479d334099daa7df374cf33c9228707aed3"},"formal_canon":{"evidence_count":2,"snapshot_sha256":"89d9492ac551897fc9375862f20ae303877ad7b91e984ff059817624ff94b4de"},"paper":{"abstract_excerpt":"Modern language models are trained almost exclusively on token sequences produced by a fixed tokenizer, an external lossless compressor often over UTF-8 byte sequences, thereby coupling the model to that compressor. This work introduces proxy compression, an alternative training scheme that preserves the efficiency benefits of compressed inputs while providing an end-to-end, raw-byte interface at inference time. During training, a single language model is jointly trained on raw byte sequences and compressed views generated by external compressors; through the process, the model learns to inter","authors_text":"Lingpeng Kong, Lin Zheng, Qian Liu, Xiachong Feng, Xinyu Li","cross_cats":["cs.LG"],"headline":"Proxy compression trains language models jointly on raw bytes and compressed sequences so they can use efficient inputs during training yet run purely on raw bytes at inference.","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2026-02-04T07:36:46Z","title":"Proxy Compression for Language Modeling"},"references":{"count":29,"internal_anchors":4,"resolved_work":29,"sample":[{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":1,"title":"Magnet: Improving the multilingual fairness of language models with adaptive gradient-based tokenization","work_id":"c19a5144-aef7-4515-a0d5-05fde00aa9e8","year":2023},{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":2,"title":"URL https://openreview.net/forum? id=PEpbUobfJv. Cao, K. and Rimell, L. You should evaluate your language model on marginal likelihood over tokeni- sations. InProceedings of the 2021 Conference on Emp","work_id":"a9c0a510-0c8c-4107-a1a5-bb766e8da0c9","year":2021},{"cited_arxiv_id":"2107.03374","doi":"","is_internal_anchor":true,"ref_index":3,"title":"Evaluating Large Language Models Trained on Code","work_id":"042493e9-b26f-4b4e-bbde-382072ca9b08","year":2021},{"cited_arxiv_id":"2407.21783","doi":"10.18653/v1/n19-1423","is_internal_anchor":true,"ref_index":4,"title":"The Llama 3 Herd of Models","work_id":"1549a635-88af-4ac1-acfe-51ae7bb53345","year":2019},{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":5,"title":"emnlp-industry.58/","work_id":"93368c2d-5c7f-41ec-a047-47b30f46d1ae","year":2023}],"snapshot_sha256":"41949df9a87d04c09e9a5d873d7ffe1991c9e5f2faee63e83e5ae8c1a4c0379b"},"source":{"id":"2602.04289","kind":"arxiv","version":2},"verdict":{"created_at":"2026-05-16T08:00:32.518982Z","id":"8f795edd-4a58-4c2c-845f-4e2ba09840a1","model_set":{"reader":"grok-4.3"},"one_line_summary":"Proxy compression trains language models on both raw bytes and compressed sequences to enable efficient training with raw-byte inference at test time.","pipeline_version":"pith-pipeline@v0.9.0","pith_extraction_headline":"Proxy compression trains language models jointly on raw bytes and compressed sequences so they can use efficient inputs during training yet run purely on raw bytes at inference.","strongest_claim":"proxy compression substantially improves training efficiency and significantly outperforms pure byte-level baselines given fixed compute budgets. As model scale increases, these gains become more pronounced, and proxy-trained models eventually match or surpass tokenizer approaches, all while operating solely on raw bytes.","weakest_assumption":"that the internal alignment learned during joint training on compressed and raw views transfers effectively to pure raw-byte inference without performance degradation or the need for continued compressed inputs."}},"verdict_id":"8f795edd-4a58-4c2c-845f-4e2ba09840a1"}}],"author_attestations":[],"timestamp_anchors":[],"storage_attestations":[],"citation_signatures":[],"replication_records":[],"corrections":[],"mirror_hints":[],"record_created":{"event_id":"sha256:13e6484db9366ab406f48f24bbd3c29a684903fe7b626eb6faaae19f325e15ed","target":"record","created_at":"2026-05-17T23:39:00Z","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":"a4eb8756ec90079c47ff7deb3085703cf3dc7bc4f068cec585aad1704d97d8c9","cross_cats_sorted":["cs.LG"],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2026-02-04T07:36:46Z","title_canon_sha256":"e920092daf9b8b6cd485968f985685df05792529c32cedad61380bddef078911"},"schema_version":"1.0","source":{"id":"2602.04289","kind":"arxiv","version":2}},"canonical_sha256":"faaba9df9231ac75f34641ecc475ebdbe92a02fce573c3b2b23b5b3b73d1c5ce","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"faaba9df9231ac75f34641ecc475ebdbe92a02fce573c3b2b23b5b3b73d1c5ce","first_computed_at":"2026-05-17T23:39:00.067532Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-05-17T23:39:00.067532Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"YHtHcDkr21kqW7U+OZ0ag+2p8YyqHVpjbWdTvkWK5GmLQFLcIiL3Wdn41MS+hPeVBTjDjKoaXnTw6N1ziU9lDA==","signature_status":"signed_v1","signed_at":"2026-05-17T23:39:00.068266Z","signed_message":"canonical_sha256_bytes"},"source_id":"2602.04289","source_kind":"arxiv","source_version":2}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:13e6484db9366ab406f48f24bbd3c29a684903fe7b626eb6faaae19f325e15ed","sha256:27206fbfa882b7306a4b048806e08e55347875247116ff7a813df833a112f45b"],"state_sha256":"658ab941c3d75f165ce00d015e6b8f11c0a3b58d2bf67951f9a7c107c361c9c9"}