{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:QJRIA2CDMFPEJPKU7NX2M2PHTL","short_pith_number":"pith:QJRIA2CD","schema_version":"1.0","canonical_sha256":"8262806843615e44bd54fb6fa669e79ac793ba4120dc025d5bf3c42334677fa3","source":{"kind":"arxiv","id":"2608.03276","version":1},"attestation_state":"computed","paper":{"title":"TaskPress: Query-Agnostic KV Cache Compression via Task-Guided Pruning","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.AI","authors_text":"Seung-won Hwang, Wonpyo Park","submitted_at":"2026-08-04T07:51:37Z","abstract_excerpt":"Long-context inference with large language models is constrained by the linear growth of the key-value cache to sequence length. While pruning offers mitigation, prevailing methods determine query-specific token importance that cannot be reused across unseen queries. In contrast, we introduce TaskPress, a framework for task-guided, query-agnostic KV cache eviction. Instead of optimizing the cache for a single query, TaskPress constructs a reusable memory representation conditioned on a high-level task guide. The guide functions as a meta-query during prefill to filter irrelevant tokens before "},"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":"2608.03276","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.AI","submitted_at":"2026-08-04T07:51:37Z","cross_cats_sorted":[],"title_canon_sha256":"73f3e455d4a00bb615d4dc267b6e0a8743fc5f025ff27341621df6287e6ad16c","abstract_canon_sha256":"ea6ec27b3b01ad484ed0450f4f5e59482cec9a41c8891ee35ab0cbb814aa3a03"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-08-05T00:46:37.576976Z","signature_b64":"J7ECtP5UiMk+xeSjKV2DMN1TAkuwcWIjSobTp0zpuAifjUwNsV+vBMrmxXQ1JvParKFpusvxa6yir9D/R07/Bg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"8262806843615e44bd54fb6fa669e79ac793ba4120dc025d5bf3c42334677fa3","last_reissued_at":"2026-08-05T00:46:37.574626Z","signature_status":"signed_v1","first_computed_at":"2026-08-05T00:46:37.574626Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"TaskPress: Query-Agnostic KV Cache Compression via Task-Guided Pruning","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.AI","authors_text":"Seung-won Hwang, Wonpyo Park","submitted_at":"2026-08-04T07:51:37Z","abstract_excerpt":"Long-context inference with large language models is constrained by the linear growth of the key-value cache to sequence length. While pruning offers mitigation, prevailing methods determine query-specific token importance that cannot be reused across unseen queries. In contrast, we introduce TaskPress, a framework for task-guided, query-agnostic KV cache eviction. Instead of optimizing the cache for a single query, TaskPress constructs a reusable memory representation conditioned on a high-level task guide. The guide functions as a meta-query during prefill to filter irrelevant tokens before "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2608.03276","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/2608.03276/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":"2608.03276","created_at":"2026-08-05T00:46:37.575442+00:00"},{"alias_kind":"arxiv_version","alias_value":"2608.03276v1","created_at":"2026-08-05T00:46:37.575442+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2608.03276","created_at":"2026-08-05T00:46:37.575442+00:00"},{"alias_kind":"pith_short_12","alias_value":"QJRIA2CDMFPE","created_at":"2026-08-05T00:46:37.575442+00:00"},{"alias_kind":"pith_short_16","alias_value":"QJRIA2CDMFPEJPKU","created_at":"2026-08-05T00:46:37.575442+00:00"},{"alias_kind":"pith_short_8","alias_value":"QJRIA2CD","created_at":"2026-08-05T00:46:37.575442+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":0,"internal_anchor_count":0,"sample":[]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/QJRIA2CDMFPEJPKU7NX2M2PHTL","json":"https://pith.science/pith/QJRIA2CDMFPEJPKU7NX2M2PHTL.json","graph_json":"https://pith.science/api/pith-number/QJRIA2CDMFPEJPKU7NX2M2PHTL/graph.json","events_json":"https://pith.science/api/pith-number/QJRIA2CDMFPEJPKU7NX2M2PHTL/events.json","paper":"https://pith.science/paper/QJRIA2CD"},"agent_actions":{"view_html":"https://pith.science/pith/QJRIA2CDMFPEJPKU7NX2M2PHTL","download_json":"https://pith.science/pith/QJRIA2CDMFPEJPKU7NX2M2PHTL.json","view_paper":"https://pith.science/paper/QJRIA2CD","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2608.03276&json=true","fetch_graph":"https://pith.science/api/pith-number/QJRIA2CDMFPEJPKU7NX2M2PHTL/graph.json","fetch_events":"https://pith.science/api/pith-number/QJRIA2CDMFPEJPKU7NX2M2PHTL/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/QJRIA2CDMFPEJPKU7NX2M2PHTL/action/timestamp_anchor","attest_storage":"https://pith.science/pith/QJRIA2CDMFPEJPKU7NX2M2PHTL/action/storage_attestation","attest_author":"https://pith.science/pith/QJRIA2CDMFPEJPKU7NX2M2PHTL/action/author_attestation","sign_citation":"https://pith.science/pith/QJRIA2CDMFPEJPKU7NX2M2PHTL/action/citation_signature","submit_replication":"https://pith.science/pith/QJRIA2CDMFPEJPKU7NX2M2PHTL/action/replication_record"}},"created_at":"2026-08-05T00:46:37.575442+00:00","updated_at":"2026-08-05T00:46:37.575442+00:00"}