{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:QQ3JJRYRWBC4CG6Q7SAT5T5K6Q","short_pith_number":"pith:QQ3JJRYR","schema_version":"1.0","canonical_sha256":"843694c711b045c11bd0fc813ecfaaf41fda2798b7cc7aa1f8aff854ad7f1705","source":{"kind":"arxiv","id":"2410.04468","version":4},"attestation_state":"computed","paper":{"title":"Revisiting In-context Learning Inference Circuit in Large Language Models","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","cs.LG"],"primary_cat":"cs.CL","authors_text":"Hakaze Cho, Mariko Kato, Naoya Inoue, Yoshihiro Sakai","submitted_at":"2024-10-06T12:50:15Z","abstract_excerpt":"In-context Learning (ICL) is an emerging few-shot learning paradigm on Language Models (LMs) with inner mechanisms un-explored. There are already existing works describing the inner processing of ICL, while they struggle to capture all the inference phenomena in large language models. Therefore, this paper proposes a comprehensive circuit to model the inference dynamics and try to explain the observed phenomena of ICL. In detail, we divide ICL inference into 3 major operations: (1) Input Text Encode: LMs encode every input text (in the demonstrations and queries) into linear representation in "},"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":"2410.04468","kind":"arxiv","version":4},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2024-10-06T12:50:15Z","cross_cats_sorted":["cs.AI","cs.LG"],"title_canon_sha256":"dfdcd9d35d3ae5178c194a1b8b282fa26a49fabb672d76372474573fb088a646","abstract_canon_sha256":"e8232f5cf877455234ac2960716d3333e37c91616f8dcf6db1d6dbd1f39a3e55"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T10:17:11.932262Z","signature_b64":"+1ewv24FoHZhcj/hZQ/dVGxP0K3SfelkNEedS+Kgp6NiHuRqhTSZ9B9YDvcU2yJMzBzLqQmHxNmelCpsBas+BQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"843694c711b045c11bd0fc813ecfaaf41fda2798b7cc7aa1f8aff854ad7f1705","last_reissued_at":"2026-07-05T10:17:11.931841Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T10:17:11.931841Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Revisiting In-context Learning Inference Circuit in Large Language Models","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","cs.LG"],"primary_cat":"cs.CL","authors_text":"Hakaze Cho, Mariko Kato, Naoya Inoue, Yoshihiro Sakai","submitted_at":"2024-10-06T12:50:15Z","abstract_excerpt":"In-context Learning (ICL) is an emerging few-shot learning paradigm on Language Models (LMs) with inner mechanisms un-explored. There are already existing works describing the inner processing of ICL, while they struggle to capture all the inference phenomena in large language models. Therefore, this paper proposes a comprehensive circuit to model the inference dynamics and try to explain the observed phenomena of ICL. In detail, we divide ICL inference into 3 major operations: (1) Input Text Encode: LMs encode every input text (in the demonstrations and queries) into linear representation in "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2410.04468","kind":"arxiv","version":4},"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/2410.04468/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":"2410.04468","created_at":"2026-07-05T10:17:11.931894+00:00"},{"alias_kind":"arxiv_version","alias_value":"2410.04468v4","created_at":"2026-07-05T10:17:11.931894+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2410.04468","created_at":"2026-07-05T10:17:11.931894+00:00"},{"alias_kind":"pith_short_12","alias_value":"QQ3JJRYRWBC4","created_at":"2026-07-05T10:17:11.931894+00:00"},{"alias_kind":"pith_short_16","alias_value":"QQ3JJRYRWBC4CG6Q","created_at":"2026-07-05T10:17:11.931894+00:00"},{"alias_kind":"pith_short_8","alias_value":"QQ3JJRYR","created_at":"2026-07-05T10:17:11.931894+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":2,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2604.08970","citing_title":"Litmus (Re)Agent: A Benchmark and Agentic System for Predictive Evaluation of Multilingual Models","ref_index":1,"is_internal_anchor":false},{"citing_arxiv_id":"2604.13403","citing_title":"Why Multimodal In-Context Learning Lags Behind? Unveiling the Inner Mechanisms and Bottlenecks","ref_index":8,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/QQ3JJRYRWBC4CG6Q7SAT5T5K6Q","json":"https://pith.science/pith/QQ3JJRYRWBC4CG6Q7SAT5T5K6Q.json","graph_json":"https://pith.science/api/pith-number/QQ3JJRYRWBC4CG6Q7SAT5T5K6Q/graph.json","events_json":"https://pith.science/api/pith-number/QQ3JJRYRWBC4CG6Q7SAT5T5K6Q/events.json","paper":"https://pith.science/paper/QQ3JJRYR"},"agent_actions":{"view_html":"https://pith.science/pith/QQ3JJRYRWBC4CG6Q7SAT5T5K6Q","download_json":"https://pith.science/pith/QQ3JJRYRWBC4CG6Q7SAT5T5K6Q.json","view_paper":"https://pith.science/paper/QQ3JJRYR","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2410.04468&json=true","fetch_graph":"https://pith.science/api/pith-number/QQ3JJRYRWBC4CG6Q7SAT5T5K6Q/graph.json","fetch_events":"https://pith.science/api/pith-number/QQ3JJRYRWBC4CG6Q7SAT5T5K6Q/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/QQ3JJRYRWBC4CG6Q7SAT5T5K6Q/action/timestamp_anchor","attest_storage":"https://pith.science/pith/QQ3JJRYRWBC4CG6Q7SAT5T5K6Q/action/storage_attestation","attest_author":"https://pith.science/pith/QQ3JJRYRWBC4CG6Q7SAT5T5K6Q/action/author_attestation","sign_citation":"https://pith.science/pith/QQ3JJRYRWBC4CG6Q7SAT5T5K6Q/action/citation_signature","submit_replication":"https://pith.science/pith/QQ3JJRYRWBC4CG6Q7SAT5T5K6Q/action/replication_record"}},"created_at":"2026-07-05T10:17:11.931894+00:00","updated_at":"2026-07-05T10:17:11.931894+00:00"}