{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2023:XQWYELDCHJ6OFVN7YQLO3OV6D4","short_pith_number":"pith:XQWYELDC","schema_version":"1.0","canonical_sha256":"bc2d822c623a7ce2d5bfc416edbabe1f334129cc04a9a4987125e82a943b5fd2","source":{"kind":"arxiv","id":"2304.08467","version":3},"attestation_state":"computed","paper":{"title":"Learning to Compress Prompts with Gist Tokens","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Jesse Mu, Noah Goodman, Xiang Lisa Li","submitted_at":"2023-04-17T17:47:37Z","abstract_excerpt":"Prompting is the primary way to utilize the multitask capabilities of language models (LMs), but prompts occupy valuable space in the input context window, and repeatedly encoding the same prompt is computationally inefficient. Finetuning and distillation methods allow for specialization of LMs without prompting, but require retraining the model for each task. To avoid this trade-off entirely, we present gisting, which trains an LM to compress prompts into smaller sets of \"gist\" tokens which can be cached and reused for compute efficiency. Gist models can be trained with no additional cost ove"},"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":"2304.08467","kind":"arxiv","version":3},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2023-04-17T17:47:37Z","cross_cats_sorted":[],"title_canon_sha256":"f6bdc7b76c8e12597045f1f3153e190ea96956942185d825f1a0eb48989002c7","abstract_canon_sha256":"b6e1ea7ee0c33e3cf436f4c90395994ff53ae6261d4f545a1b971c2084f76c20"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T07:44:17.726858Z","signature_b64":"LWZRp0L8wHWnTlu0Tx/9L3v/byEdEAwdjLq/cchqtWVmZ/PsBapF1hgpljP4iNfrn8Nlad2OFxE6C0Sc2Mm6BQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"bc2d822c623a7ce2d5bfc416edbabe1f334129cc04a9a4987125e82a943b5fd2","last_reissued_at":"2026-07-05T07:44:17.726421Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T07:44:17.726421Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Learning to Compress Prompts with Gist Tokens","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Jesse Mu, Noah Goodman, Xiang Lisa Li","submitted_at":"2023-04-17T17:47:37Z","abstract_excerpt":"Prompting is the primary way to utilize the multitask capabilities of language models (LMs), but prompts occupy valuable space in the input context window, and repeatedly encoding the same prompt is computationally inefficient. Finetuning and distillation methods allow for specialization of LMs without prompting, but require retraining the model for each task. To avoid this trade-off entirely, we present gisting, which trains an LM to compress prompts into smaller sets of \"gist\" tokens which can be cached and reused for compute efficiency. Gist models can be trained with no additional cost ove"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2304.08467","kind":"arxiv","version":3},"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/2304.08467/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":"2304.08467","created_at":"2026-07-05T07:44:17.726479+00:00"},{"alias_kind":"arxiv_version","alias_value":"2304.08467v3","created_at":"2026-07-05T07:44:17.726479+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2304.08467","created_at":"2026-07-05T07:44:17.726479+00:00"},{"alias_kind":"pith_short_12","alias_value":"XQWYELDCHJ6O","created_at":"2026-07-05T07:44:17.726479+00:00"},{"alias_kind":"pith_short_16","alias_value":"XQWYELDCHJ6OFVN7","created_at":"2026-07-05T07:44:17.726479+00:00"},{"alias_kind":"pith_short_8","alias_value":"XQWYELDC","created_at":"2026-07-05T07:44:17.726479+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":8,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.20295","citing_title":"Token-Operations-Oriented Inference Optimization Techniques for Large Models","ref_index":125,"is_internal_anchor":false},{"citing_arxiv_id":"2605.22337","citing_title":"Meta-Soft: Leveraging Composable Meta-Tokens for Context-Preserving KV Cache Compression","ref_index":19,"is_internal_anchor":false},{"citing_arxiv_id":"2605.22337","citing_title":"Meta-Soft: Leveraging Composable Meta-Tokens for Context-Preserving KV Cache Compression","ref_index":19,"is_internal_anchor":false},{"citing_arxiv_id":"2306.14048","citing_title":"H$_2$O: Heavy-Hitter Oracle for Efficient Generative Inference of Large Language Models","ref_index":14,"is_internal_anchor":false},{"citing_arxiv_id":"2310.01801","citing_title":"Model Tells You What to Discard: Adaptive KV Cache Compression for LLMs","ref_index":91,"is_internal_anchor":false},{"citing_arxiv_id":"2404.14294","citing_title":"A Survey on Efficient Inference for Large Language Models","ref_index":34,"is_internal_anchor":false},{"citing_arxiv_id":"2605.04107","citing_title":"TSCG: Deterministic Tool-Schema Compilation for Agentic LLM Deployments","ref_index":15,"is_internal_anchor":false},{"citing_arxiv_id":"2604.13725","citing_title":"On the Effectiveness of Context Compression for Repository-Level Tasks: An Empirical Investigation","ref_index":25,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/XQWYELDCHJ6OFVN7YQLO3OV6D4","json":"https://pith.science/pith/XQWYELDCHJ6OFVN7YQLO3OV6D4.json","graph_json":"https://pith.science/api/pith-number/XQWYELDCHJ6OFVN7YQLO3OV6D4/graph.json","events_json":"https://pith.science/api/pith-number/XQWYELDCHJ6OFVN7YQLO3OV6D4/events.json","paper":"https://pith.science/paper/XQWYELDC"},"agent_actions":{"view_html":"https://pith.science/pith/XQWYELDCHJ6OFVN7YQLO3OV6D4","download_json":"https://pith.science/pith/XQWYELDCHJ6OFVN7YQLO3OV6D4.json","view_paper":"https://pith.science/paper/XQWYELDC","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2304.08467&json=true","fetch_graph":"https://pith.science/api/pith-number/XQWYELDCHJ6OFVN7YQLO3OV6D4/graph.json","fetch_events":"https://pith.science/api/pith-number/XQWYELDCHJ6OFVN7YQLO3OV6D4/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/XQWYELDCHJ6OFVN7YQLO3OV6D4/action/timestamp_anchor","attest_storage":"https://pith.science/pith/XQWYELDCHJ6OFVN7YQLO3OV6D4/action/storage_attestation","attest_author":"https://pith.science/pith/XQWYELDCHJ6OFVN7YQLO3OV6D4/action/author_attestation","sign_citation":"https://pith.science/pith/XQWYELDCHJ6OFVN7YQLO3OV6D4/action/citation_signature","submit_replication":"https://pith.science/pith/XQWYELDCHJ6OFVN7YQLO3OV6D4/action/replication_record"}},"created_at":"2026-07-05T07:44:17.726479+00:00","updated_at":"2026-07-05T07:44:17.726479+00:00"}