{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:M5VNWOCU5TPKCTD45KX2J4TCF7","short_pith_number":"pith:M5VNWOCU","schema_version":"1.0","canonical_sha256":"676adb3854ecdea14c7ceaafa4f2622fd59538be4e956b68fc67b62d4bd9ee20","source":{"kind":"arxiv","id":"2510.20535","version":2},"attestation_state":"computed","paper":{"title":"ARC-Encoder: learning compressed text representations for large language models","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CL","authors_text":"Edouard Grave, Hippolyte Pilchen, Patrick P\\'erez","submitted_at":"2025-10-23T13:20:57Z","abstract_excerpt":"Recent techniques such as retrieval-augmented generation or chain-of-thought reasoning have led to longer contexts and increased inference costs. Context compression techniques can reduce these costs, but the most effective approaches require fine-tuning the target model or even modifying its architecture. This can degrade its general abilities when not used for this specific purpose. Here we explore an alternative approach: an encoder that compresses the context into continuous representations which replace token embeddings in decoder LLMs. First, we perform a systematic study of training str"},"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":false,"formal_links_present":false},"canonical_record":{"source":{"id":"2510.20535","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2025-10-23T13:20:57Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"3714835707f9c7f78e4f216785ad9b5941f6db0aa70d07a6a899df2192636d0e","abstract_canon_sha256":"87b3e57bc9b2e368ca6b372b286e041dec6d86aabbba29d4f594e970a244459b"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"676adb3854ecdea14c7ceaafa4f2622fd59538be4e956b68fc67b62d4bd9ee20","last_reissued_at":"2026-07-30T01:19:14.822044Z","signature_status":"unsigned_v0","first_computed_at":"2026-07-30T01:19:14.822044Z"},"graph_snapshot":{"paper":{"title":"ARC-Encoder: learning compressed text representations for large language models","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CL","authors_text":"Edouard Grave, Hippolyte Pilchen, Patrick P\\'erez","submitted_at":"2025-10-23T13:20:57Z","abstract_excerpt":"Recent techniques such as retrieval-augmented generation or chain-of-thought reasoning have led to longer contexts and increased inference costs. Context compression techniques can reduce these costs, but the most effective approaches require fine-tuning the target model or even modifying its architecture. This can degrade its general abilities when not used for this specific purpose. Here we explore an alternative approach: an encoder that compresses the context into continuous representations which replace token embeddings in decoder LLMs. First, we perform a systematic study of training str"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2510.20535","kind":"arxiv","version":2},"verdict":{"id":null,"model_set":{},"created_at":null,"strongest_claim":"","one_line_summary":"","pipeline_version":null,"weakest_assumption":"","pith_extraction_headline":""},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2510.20535/integrity.json","findings":[],"available":true,"detectors_run":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938"},"references":{"count":0,"sample":[],"resolved_work":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57","internal_anchors":0},"formal_canon":{"evidence_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"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":"2510.20535","created_at":"2026-07-30T01:19:14.827239+00:00"},{"alias_kind":"arxiv_version","alias_value":"2510.20535v2","created_at":"2026-07-30T01:19:14.827239+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2510.20535","created_at":"2026-07-30T01:19:14.827239+00:00"},{"alias_kind":"pith_short_12","alias_value":"M5VNWOCU5TPK","created_at":"2026-07-30T01:19:14.827239+00:00"},{"alias_kind":"pith_short_16","alias_value":"M5VNWOCU5TPKCTD4","created_at":"2026-07-30T01:19:14.827239+00:00"},{"alias_kind":"pith_short_8","alias_value":"M5VNWOCU","created_at":"2026-07-30T01:19:14.827239+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":2,"internal_anchor_count":2,"sample":[{"citing_arxiv_id":"2606.09659","citing_title":"End-to-End Context Compression at Scale","ref_index":71,"is_internal_anchor":true},{"citing_arxiv_id":"2604.26483","citing_title":"Efficient Listwise Reranking with Compressed Document Representations","ref_index":24,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/M5VNWOCU5TPKCTD45KX2J4TCF7","json":"https://pith.science/pith/M5VNWOCU5TPKCTD45KX2J4TCF7.json","graph_json":"https://pith.science/api/pith-number/M5VNWOCU5TPKCTD45KX2J4TCF7/graph.json","events_json":"https://pith.science/api/pith-number/M5VNWOCU5TPKCTD45KX2J4TCF7/events.json","paper":"https://pith.science/paper/M5VNWOCU"},"agent_actions":{"view_html":"https://pith.science/pith/M5VNWOCU5TPKCTD45KX2J4TCF7","download_json":"https://pith.science/pith/M5VNWOCU5TPKCTD45KX2J4TCF7.json","view_paper":"https://pith.science/paper/M5VNWOCU","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2510.20535&json=true","fetch_graph":"https://pith.science/api/pith-number/M5VNWOCU5TPKCTD45KX2J4TCF7/graph.json","fetch_events":"https://pith.science/api/pith-number/M5VNWOCU5TPKCTD45KX2J4TCF7/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/M5VNWOCU5TPKCTD45KX2J4TCF7/action/timestamp_anchor","attest_storage":"https://pith.science/pith/M5VNWOCU5TPKCTD45KX2J4TCF7/action/storage_attestation","attest_author":"https://pith.science/pith/M5VNWOCU5TPKCTD45KX2J4TCF7/action/author_attestation","sign_citation":"https://pith.science/pith/M5VNWOCU5TPKCTD45KX2J4TCF7/action/citation_signature","submit_replication":"https://pith.science/pith/M5VNWOCU5TPKCTD45KX2J4TCF7/action/replication_record"}},"created_at":"2026-07-30T01:19:14.827239+00:00","updated_at":"2026-07-30T01:19:14.827239+00:00"}