{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:CPOHFUVYRUPL2X7SQBFY47QFUD","short_pith_number":"pith:CPOHFUVY","schema_version":"1.0","canonical_sha256":"13dc72d2b88d1ebd5ff2804b8e7e05a0e83a991b625b6b1a372da4822e14525a","source":{"kind":"arxiv","id":"2406.16254","version":2},"attestation_state":"computed","paper":{"title":"Confidence Regulation Neurons in Language Models","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.CL"],"primary_cat":"cs.LG","authors_text":"Alessandro Stolfo, Ben Wu, Mrinmaya Sachan, Neel Nanda, Wes Gurnee, Xingyi Song, Yonatan Belinkov","submitted_at":"2024-06-24T01:31:03Z","abstract_excerpt":"Despite their widespread use, the mechanisms by which large language models (LLMs) represent and regulate uncertainty in next-token predictions remain largely unexplored. This study investigates two critical components believed to influence this uncertainty: the recently discovered entropy neurons and a new set of components that we term token frequency neurons. Entropy neurons are characterized by an unusually high weight norm and influence the final layer normalization (LayerNorm) scale to effectively scale down the logits. Our work shows that entropy neurons operate by writing onto an unemb"},"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":"2406.16254","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2024-06-24T01:31:03Z","cross_cats_sorted":["cs.AI","cs.CL"],"title_canon_sha256":"307af5579f33e45e8dbd8d5ed670dcc3224999af025029a9d94667a38566cbf2","abstract_canon_sha256":"4a545b8c8f5b4763e254b36b5a4348ec76723f5cb467c4488f2caafd4251d655"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T09:32:45.380052Z","signature_b64":"cN1TP/3q+ZX9LdYwIG9AmMHXVcnqWzqKOt9f/RDmrhIazbAwNqfQjx+xrD/hInh9AEBaP2r/qIPuSd0o+tTPCQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"13dc72d2b88d1ebd5ff2804b8e7e05a0e83a991b625b6b1a372da4822e14525a","last_reissued_at":"2026-07-05T09:32:45.379078Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T09:32:45.379078Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Confidence Regulation Neurons in Language Models","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.CL"],"primary_cat":"cs.LG","authors_text":"Alessandro Stolfo, Ben Wu, Mrinmaya Sachan, Neel Nanda, Wes Gurnee, Xingyi Song, Yonatan Belinkov","submitted_at":"2024-06-24T01:31:03Z","abstract_excerpt":"Despite their widespread use, the mechanisms by which large language models (LLMs) represent and regulate uncertainty in next-token predictions remain largely unexplored. This study investigates two critical components believed to influence this uncertainty: the recently discovered entropy neurons and a new set of components that we term token frequency neurons. Entropy neurons are characterized by an unusually high weight norm and influence the final layer normalization (LayerNorm) scale to effectively scale down the logits. Our work shows that entropy neurons operate by writing onto an unemb"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2406.16254","kind":"arxiv","version":2},"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/2406.16254/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":"2406.16254","created_at":"2026-07-05T09:32:45.379588+00:00"},{"alias_kind":"arxiv_version","alias_value":"2406.16254v2","created_at":"2026-07-05T09:32:45.379588+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2406.16254","created_at":"2026-07-05T09:32:45.379588+00:00"},{"alias_kind":"pith_short_12","alias_value":"CPOHFUVYRUPL","created_at":"2026-07-05T09:32:45.379588+00:00"},{"alias_kind":"pith_short_16","alias_value":"CPOHFUVYRUPL2X7S","created_at":"2026-07-05T09:32:45.379588+00:00"},{"alias_kind":"pith_short_8","alias_value":"CPOHFUVY","created_at":"2026-07-05T09:32:45.379588+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":3,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.20743","citing_title":"Massive Activations Are Architecturally Robust: A Controlled Scratch/Commitment Residual Stream Test","ref_index":4,"is_internal_anchor":false},{"citing_arxiv_id":"2604.07098","citing_title":"Selective Neuron Amplification in Transformer Language Models","ref_index":9,"is_internal_anchor":false},{"citing_arxiv_id":"2604.07098","citing_title":"Selective Neuron Amplification in Transformer Language Models","ref_index":9,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/CPOHFUVYRUPL2X7SQBFY47QFUD","json":"https://pith.science/pith/CPOHFUVYRUPL2X7SQBFY47QFUD.json","graph_json":"https://pith.science/api/pith-number/CPOHFUVYRUPL2X7SQBFY47QFUD/graph.json","events_json":"https://pith.science/api/pith-number/CPOHFUVYRUPL2X7SQBFY47QFUD/events.json","paper":"https://pith.science/paper/CPOHFUVY"},"agent_actions":{"view_html":"https://pith.science/pith/CPOHFUVYRUPL2X7SQBFY47QFUD","download_json":"https://pith.science/pith/CPOHFUVYRUPL2X7SQBFY47QFUD.json","view_paper":"https://pith.science/paper/CPOHFUVY","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2406.16254&json=true","fetch_graph":"https://pith.science/api/pith-number/CPOHFUVYRUPL2X7SQBFY47QFUD/graph.json","fetch_events":"https://pith.science/api/pith-number/CPOHFUVYRUPL2X7SQBFY47QFUD/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/CPOHFUVYRUPL2X7SQBFY47QFUD/action/timestamp_anchor","attest_storage":"https://pith.science/pith/CPOHFUVYRUPL2X7SQBFY47QFUD/action/storage_attestation","attest_author":"https://pith.science/pith/CPOHFUVYRUPL2X7SQBFY47QFUD/action/author_attestation","sign_citation":"https://pith.science/pith/CPOHFUVYRUPL2X7SQBFY47QFUD/action/citation_signature","submit_replication":"https://pith.science/pith/CPOHFUVYRUPL2X7SQBFY47QFUD/action/replication_record"}},"created_at":"2026-07-05T09:32:45.379588+00:00","updated_at":"2026-07-05T09:32:45.379588+00:00"}