{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:73Y63LUYAPJK45AYNAISW3XV3R","short_pith_number":"pith:73Y63LUY","schema_version":"1.0","canonical_sha256":"fef1edae9803d2ae741868112b6ef5dc6120885893faa5b98fac2d72465b581d","source":{"kind":"arxiv","id":"2504.08934","version":1},"attestation_state":"computed","paper":{"title":"Long Context In-Context Compression by Getting to the Gist of Gisting","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.LG","authors_text":"Aleksandar Petrov, Andrey Zhmoginov, Mark Sandler, Max Vladymyrov, Nolan Miller","submitted_at":"2025-04-11T19:23:31Z","abstract_excerpt":"Long context processing is critical for the adoption of LLMs, but existing methods often introduce architectural complexity that hinders their practical adoption. Gisting, an in-context compression method with no architectural modification to the decoder transformer, is a promising approach due to its simplicity and compatibility with existing frameworks. While effective for short instructions, we demonstrate that gisting struggles with longer contexts, with significant performance drops even at minimal compression rates. Surprisingly, a simple average pooling baseline consistently outperforms"},"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":"2504.08934","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2025-04-11T19:23:31Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"a7456443b073fa875d5d4b49f251e5e1f9f86fd7122266723369b93e3deccbe2","abstract_canon_sha256":"be939f1fd04b8e6ee4800a04af273cc8e07a54a68ff72ba7aab8d3598d19d257"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T10:48:01.828661Z","signature_b64":"/gom32V63P6XpXF2nHT5CqfPNMW/jrkeV28AFBIZg1SteS5loZdfE8SU+gAFPuF73AVREdLizHPh8HXoVu9EBg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"fef1edae9803d2ae741868112b6ef5dc6120885893faa5b98fac2d72465b581d","last_reissued_at":"2026-07-05T10:48:01.828183Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T10:48:01.828183Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Long Context In-Context Compression by Getting to the Gist of Gisting","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.LG","authors_text":"Aleksandar Petrov, Andrey Zhmoginov, Mark Sandler, Max Vladymyrov, Nolan Miller","submitted_at":"2025-04-11T19:23:31Z","abstract_excerpt":"Long context processing is critical for the adoption of LLMs, but existing methods often introduce architectural complexity that hinders their practical adoption. Gisting, an in-context compression method with no architectural modification to the decoder transformer, is a promising approach due to its simplicity and compatibility with existing frameworks. While effective for short instructions, we demonstrate that gisting struggles with longer contexts, with significant performance drops even at minimal compression rates. Surprisingly, a simple average pooling baseline consistently outperforms"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2504.08934","kind":"arxiv","version":1},"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/2504.08934/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":"2504.08934","created_at":"2026-07-05T10:48:01.828239+00:00"},{"alias_kind":"arxiv_version","alias_value":"2504.08934v1","created_at":"2026-07-05T10:48:01.828239+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2504.08934","created_at":"2026-07-05T10:48:01.828239+00:00"},{"alias_kind":"pith_short_12","alias_value":"73Y63LUYAPJK","created_at":"2026-07-05T10:48:01.828239+00:00"},{"alias_kind":"pith_short_16","alias_value":"73Y63LUYAPJK45AY","created_at":"2026-07-05T10:48:01.828239+00:00"},{"alias_kind":"pith_short_8","alias_value":"73Y63LUY","created_at":"2026-07-05T10:48:01.828239+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":2,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.13316","citing_title":"ReSum: Synergizing LLM Reasoning and Summarization with Reinforcement Learning","ref_index":51,"is_internal_anchor":false},{"citing_arxiv_id":"2604.20920","citing_title":"Simplified Sparse Attention via Gist Tokens","ref_index":25,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/73Y63LUYAPJK45AYNAISW3XV3R","json":"https://pith.science/pith/73Y63LUYAPJK45AYNAISW3XV3R.json","graph_json":"https://pith.science/api/pith-number/73Y63LUYAPJK45AYNAISW3XV3R/graph.json","events_json":"https://pith.science/api/pith-number/73Y63LUYAPJK45AYNAISW3XV3R/events.json","paper":"https://pith.science/paper/73Y63LUY"},"agent_actions":{"view_html":"https://pith.science/pith/73Y63LUYAPJK45AYNAISW3XV3R","download_json":"https://pith.science/pith/73Y63LUYAPJK45AYNAISW3XV3R.json","view_paper":"https://pith.science/paper/73Y63LUY","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2504.08934&json=true","fetch_graph":"https://pith.science/api/pith-number/73Y63LUYAPJK45AYNAISW3XV3R/graph.json","fetch_events":"https://pith.science/api/pith-number/73Y63LUYAPJK45AYNAISW3XV3R/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/73Y63LUYAPJK45AYNAISW3XV3R/action/timestamp_anchor","attest_storage":"https://pith.science/pith/73Y63LUYAPJK45AYNAISW3XV3R/action/storage_attestation","attest_author":"https://pith.science/pith/73Y63LUYAPJK45AYNAISW3XV3R/action/author_attestation","sign_citation":"https://pith.science/pith/73Y63LUYAPJK45AYNAISW3XV3R/action/citation_signature","submit_replication":"https://pith.science/pith/73Y63LUYAPJK45AYNAISW3XV3R/action/replication_record"}},"created_at":"2026-07-05T10:48:01.828239+00:00","updated_at":"2026-07-05T10:48:01.828239+00:00"}