{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:RGQM4TQFK4TOECWO2SFCX6ICTK","short_pith_number":"pith:RGQM4TQF","schema_version":"1.0","canonical_sha256":"89a0ce4e055726e20aced48a2bf9029aa5c45a651a2a1cef16607265caa6403c","source":{"kind":"arxiv","id":"2402.11004","version":1},"attestation_state":"computed","paper":{"title":"The Evolution of Statistical Induction Heads: In-Context Learning Markov Chains","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Benjamin L. Edelman, Eran Malach, Ezra Edelman, Nikolaos Tsilivis, Surbhi Goel","submitted_at":"2024-02-16T18:28:36Z","abstract_excerpt":"Large language models have the ability to generate text that mimics patterns in their inputs. We introduce a simple Markov Chain sequence modeling task in order to study how this in-context learning (ICL) capability emerges. In our setting, each example is sampled from a Markov chain drawn from a prior distribution over Markov chains. Transformers trained on this task form \\emph{statistical induction heads} which compute accurate next-token probabilities given the bigram statistics of the context. During the course of training, models pass through multiple phases: after an initial stage in whi"},"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":"2402.11004","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2024-02-16T18:28:36Z","cross_cats_sorted":[],"title_canon_sha256":"3308d52a251ef76ab5e20030da5eca01cf85981df775e702ce90660f85540ba8","abstract_canon_sha256":"9ada4e7237f98d857798aef6868a1e238338b656086edb8b1ba6cb960685ff0f"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T07:46:03.392908Z","signature_b64":"5JM0gzdfUAUa1/pObosI0JgVmoMoC0e+8iMeA/GqhZYPDTicCETsiQxCZLP/yGbftkGjeJMardpQwjyFcSWOAw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"89a0ce4e055726e20aced48a2bf9029aa5c45a651a2a1cef16607265caa6403c","last_reissued_at":"2026-07-05T07:46:03.392444Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T07:46:03.392444Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"The Evolution of Statistical Induction Heads: In-Context Learning Markov Chains","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Benjamin L. Edelman, Eran Malach, Ezra Edelman, Nikolaos Tsilivis, Surbhi Goel","submitted_at":"2024-02-16T18:28:36Z","abstract_excerpt":"Large language models have the ability to generate text that mimics patterns in their inputs. We introduce a simple Markov Chain sequence modeling task in order to study how this in-context learning (ICL) capability emerges. In our setting, each example is sampled from a Markov chain drawn from a prior distribution over Markov chains. Transformers trained on this task form \\emph{statistical induction heads} which compute accurate next-token probabilities given the bigram statistics of the context. During the course of training, models pass through multiple phases: after an initial stage in whi"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2402.11004","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/2402.11004/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":"2402.11004","created_at":"2026-07-05T07:46:03.392511+00:00"},{"alias_kind":"arxiv_version","alias_value":"2402.11004v1","created_at":"2026-07-05T07:46:03.392511+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2402.11004","created_at":"2026-07-05T07:46:03.392511+00:00"},{"alias_kind":"pith_short_12","alias_value":"RGQM4TQFK4TO","created_at":"2026-07-05T07:46:03.392511+00:00"},{"alias_kind":"pith_short_16","alias_value":"RGQM4TQFK4TOECWO","created_at":"2026-07-05T07:46:03.392511+00:00"},{"alias_kind":"pith_short_8","alias_value":"RGQM4TQF","created_at":"2026-07-05T07:46:03.392511+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":2,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2503.10574","citing_title":"The LZ78 Source","ref_index":14,"is_internal_anchor":false},{"citing_arxiv_id":"2506.07298","citing_title":"Pre-trained Large Language Models Learn Hidden Markov Models In-context","ref_index":13,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/RGQM4TQFK4TOECWO2SFCX6ICTK","json":"https://pith.science/pith/RGQM4TQFK4TOECWO2SFCX6ICTK.json","graph_json":"https://pith.science/api/pith-number/RGQM4TQFK4TOECWO2SFCX6ICTK/graph.json","events_json":"https://pith.science/api/pith-number/RGQM4TQFK4TOECWO2SFCX6ICTK/events.json","paper":"https://pith.science/paper/RGQM4TQF"},"agent_actions":{"view_html":"https://pith.science/pith/RGQM4TQFK4TOECWO2SFCX6ICTK","download_json":"https://pith.science/pith/RGQM4TQFK4TOECWO2SFCX6ICTK.json","view_paper":"https://pith.science/paper/RGQM4TQF","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2402.11004&json=true","fetch_graph":"https://pith.science/api/pith-number/RGQM4TQFK4TOECWO2SFCX6ICTK/graph.json","fetch_events":"https://pith.science/api/pith-number/RGQM4TQFK4TOECWO2SFCX6ICTK/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/RGQM4TQFK4TOECWO2SFCX6ICTK/action/timestamp_anchor","attest_storage":"https://pith.science/pith/RGQM4TQFK4TOECWO2SFCX6ICTK/action/storage_attestation","attest_author":"https://pith.science/pith/RGQM4TQFK4TOECWO2SFCX6ICTK/action/author_attestation","sign_citation":"https://pith.science/pith/RGQM4TQFK4TOECWO2SFCX6ICTK/action/citation_signature","submit_replication":"https://pith.science/pith/RGQM4TQFK4TOECWO2SFCX6ICTK/action/replication_record"}},"created_at":"2026-07-05T07:46:03.392511+00:00","updated_at":"2026-07-05T07:46:03.392511+00:00"}