{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:7KV2TX4SGGWHL42GIHWMI5PL3P","short_pith_number":"pith:7KV2TX4S","schema_version":"1.0","canonical_sha256":"faaba9df9231ac75f34641ecc475ebdbe92a02fce573c3b2b23b5b3b73d1c5ce","source":{"kind":"arxiv","id":"2602.04289","version":2},"attestation_state":"computed","paper":{"title":"Proxy Compression for Language Modeling","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","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.","cross_cats":["cs.LG"],"primary_cat":"cs.CL","authors_text":"Lingpeng Kong, Lin Zheng, Qian Liu, Xiachong Feng, Xinyu Li","submitted_at":"2026-02-04T07:36:46Z","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"},"verification_status":{"content_addressed":true,"pith_receipt":true,"author_attested":false,"weak_author_claims":0,"strong_author_claims":0,"externally_anchored":false,"storage_verified":false,"citation_signatures":0,"replication_records":0,"graph_snapshot":true,"references_resolved":true,"formal_links_present":true},"canonical_record":{"source":{"id":"2602.04289","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2026-02-04T07:36:46Z","cross_cats_sorted":["cs.LG"],"title_canon_sha256":"e920092daf9b8b6cd485968f985685df05792529c32cedad61380bddef078911","abstract_canon_sha256":"a4eb8756ec90079c47ff7deb3085703cf3dc7bc4f068cec585aad1704d97d8c9"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-17T23:39:00.068266Z","signature_b64":"YHtHcDkr21kqW7U+OZ0ag+2p8YyqHVpjbWdTvkWK5GmLQFLcIiL3Wdn41MS+hPeVBTjDjKoaXnTw6N1ziU9lDA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"faaba9df9231ac75f34641ecc475ebdbe92a02fce573c3b2b23b5b3b73d1c5ce","last_reissued_at":"2026-05-17T23:39:00.067532Z","signature_status":"signed_v1","first_computed_at":"2026-05-17T23:39:00.067532Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Proxy Compression for Language Modeling","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","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.","cross_cats":["cs.LG"],"primary_cat":"cs.CL","authors_text":"Lingpeng Kong, Lin Zheng, Qian Liu, Xiachong Feng, Xinyu Li","submitted_at":"2026-02-04T07:36:46Z","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"},"claims":{"count":4,"items":[{"kind":"strongest_claim","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.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","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.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"Proxy compression trains language models on both raw bytes and compressed sequences to enable efficient training with raw-byte inference at test time.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","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.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"bd42ab1bbd133da7cccab68460c3a479d334099daa7df374cf33c9228707aed3"},"source":{"id":"2602.04289","kind":"arxiv","version":2},"verdict":{"id":"8f795edd-4a58-4c2c-845f-4e2ba09840a1","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-16T08:00:32.518982Z","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.","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","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.","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."},"references":{"count":29,"sample":[{"doi":"","year":2023,"title":"Magnet: Improving the multilingual fairness of language models with adaptive gradient-based tokenization","work_id":"c19a5144-aef7-4515-a0d5-05fde00aa9e8","ref_index":1,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2021,"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","ref_index":2,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2021,"title":"Evaluating Large Language Models Trained on Code","work_id":"042493e9-b26f-4b4e-bbde-382072ca9b08","ref_index":3,"cited_arxiv_id":"2107.03374","is_internal_anchor":true},{"doi":"10.18653/v1/n19-1423","year":2019,"title":"The Llama 3 Herd of Models","work_id":"1549a635-88af-4ac1-acfe-51ae7bb53345","ref_index":4,"cited_arxiv_id":"2407.21783","is_internal_anchor":true},{"doi":"","year":2023,"title":"emnlp-industry.58/","work_id":"93368c2d-5c7f-41ec-a047-47b30f46d1ae","ref_index":5,"cited_arxiv_id":"","is_internal_anchor":false}],"resolved_work":29,"snapshot_sha256":"41949df9a87d04c09e9a5d873d7ffe1991c9e5f2faee63e83e5ae8c1a4c0379b","internal_anchors":4},"formal_canon":{"evidence_count":2,"snapshot_sha256":"89d9492ac551897fc9375862f20ae303877ad7b91e984ff059817624ff94b4de"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"},"aliases":[{"alias_kind":"arxiv","alias_value":"2602.04289","created_at":"2026-05-17T23:39:00.067685+00:00"},{"alias_kind":"arxiv_version","alias_value":"2602.04289v2","created_at":"2026-05-17T23:39:00.067685+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2602.04289","created_at":"2026-05-17T23:39:00.067685+00:00"},{"alias_kind":"pith_short_12","alias_value":"7KV2TX4SGGWH","created_at":"2026-05-18T12:33:37.589309+00:00"},{"alias_kind":"pith_short_16","alias_value":"7KV2TX4SGGWHL42G","created_at":"2026-05-18T12:33:37.589309+00:00"},{"alias_kind":"pith_short_8","alias_value":"7KV2TX4S","created_at":"2026-05-18T12:33:37.589309+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":5,"internal_anchor_count":5,"sample":[{"citing_arxiv_id":"2605.06546","citing_title":"Efficient Pre-Training with Token Superposition","ref_index":65,"is_internal_anchor":true},{"citing_arxiv_id":"2604.27263","citing_title":"Decoupling the Benefits of Subword Tokenization for Language Model Training via Byte-level Simulation","ref_index":37,"is_internal_anchor":true},{"citing_arxiv_id":"2605.09630","citing_title":"Scratchpad Patching: Decoupling Compute from Patch Size in Byte-Level Language Models","ref_index":103,"is_internal_anchor":true},{"citing_arxiv_id":"2605.06546","citing_title":"Efficient Pre-Training with Token Superposition","ref_index":63,"is_internal_anchor":true},{"citing_arxiv_id":"2604.27263","citing_title":"Decoupling the Benefits of Subword Tokenization for Language Model Training via Byte-level Simulation","ref_index":37,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":2,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/7KV2TX4SGGWHL42GIHWMI5PL3P","json":"https://pith.science/pith/7KV2TX4SGGWHL42GIHWMI5PL3P.json","graph_json":"https://pith.science/api/pith-number/7KV2TX4SGGWHL42GIHWMI5PL3P/graph.json","events_json":"https://pith.science/api/pith-number/7KV2TX4SGGWHL42GIHWMI5PL3P/events.json","paper":"https://pith.science/paper/7KV2TX4S"},"agent_actions":{"view_html":"https://pith.science/pith/7KV2TX4SGGWHL42GIHWMI5PL3P","download_json":"https://pith.science/pith/7KV2TX4SGGWHL42GIHWMI5PL3P.json","view_paper":"https://pith.science/paper/7KV2TX4S","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2602.04289&json=true","fetch_graph":"https://pith.science/api/pith-number/7KV2TX4SGGWHL42GIHWMI5PL3P/graph.json","fetch_events":"https://pith.science/api/pith-number/7KV2TX4SGGWHL42GIHWMI5PL3P/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/7KV2TX4SGGWHL42GIHWMI5PL3P/action/timestamp_anchor","attest_storage":"https://pith.science/pith/7KV2TX4SGGWHL42GIHWMI5PL3P/action/storage_attestation","attest_author":"https://pith.science/pith/7KV2TX4SGGWHL42GIHWMI5PL3P/action/author_attestation","sign_citation":"https://pith.science/pith/7KV2TX4SGGWHL42GIHWMI5PL3P/action/citation_signature","submit_replication":"https://pith.science/pith/7KV2TX4SGGWHL42GIHWMI5PL3P/action/replication_record"}},"created_at":"2026-05-17T23:39:00.067685+00:00","updated_at":"2026-05-17T23:39:00.067685+00:00"}