{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:E2P7QHWXKHA4LDWOAPATUSOOE4","short_pith_number":"pith:E2P7QHWX","schema_version":"1.0","canonical_sha256":"269ff81ed751c1c58ece03c13a49ce272c39ccd30e41d362ee9845850a7eef21","source":{"kind":"arxiv","id":"2608.01816","version":1},"attestation_state":"computed","paper":{"title":"Divergent large language model predictions from convergent representations in ambiguous word pairs","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"K. Jack Scott, Narun Pat, Veronica Liesaputra","submitted_at":"2026-08-03T07:26:45Z","abstract_excerpt":"In this work we investigate how decoder-only transformers resolve lexical ambiguity through layer-by-layer analysis of three models spanning three parameter sizes (GPT-2-Small-117M, Llama-3.2-3B, Qwen2.5-32B). For both homonyms and polysemes, we find that representations become maximally distinct in middle layers, then partially reconverge in late layers, while the KL divergence between their next-token predictions reaches its maximum in the final layers. The activation patching experiment provides causal evidence that late-layer representational differences directly determine outputs despite "},"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":"2608.01816","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2026-08-03T07:26:45Z","cross_cats_sorted":[],"title_canon_sha256":"315ba5d04b14d56ebf066b5e6b5c0fa0e8a859ce42d49add4065228cfa1aaa40","abstract_canon_sha256":"aa1d2615f5a50da7462b7d3f9a5d4a1e6e0a0252c4409f25ce1993750df36393"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-08-04T02:07:14.094958Z","signature_b64":"cOtKOcsCicW+B0mBbh+4XIUccjJWtxaZrqIPuQ5ItA6TpEhkT+ujP2C5u1rFIj46P/jDtaNnaLfoQPGGPZRpCw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"269ff81ed751c1c58ece03c13a49ce272c39ccd30e41d362ee9845850a7eef21","last_reissued_at":"2026-08-04T02:07:14.093491Z","signature_status":"signed_v1","first_computed_at":"2026-08-04T02:07:14.093491Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Divergent large language model predictions from convergent representations in ambiguous word pairs","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"K. Jack Scott, Narun Pat, Veronica Liesaputra","submitted_at":"2026-08-03T07:26:45Z","abstract_excerpt":"In this work we investigate how decoder-only transformers resolve lexical ambiguity through layer-by-layer analysis of three models spanning three parameter sizes (GPT-2-Small-117M, Llama-3.2-3B, Qwen2.5-32B). For both homonyms and polysemes, we find that representations become maximally distinct in middle layers, then partially reconverge in late layers, while the KL divergence between their next-token predictions reaches its maximum in the final layers. The activation patching experiment provides causal evidence that late-layer representational differences directly determine outputs despite "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2608.01816","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/2608.01816/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":"2608.01816","created_at":"2026-08-04T02:07:14.094928+00:00"},{"alias_kind":"arxiv_version","alias_value":"2608.01816v1","created_at":"2026-08-04T02:07:14.094928+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2608.01816","created_at":"2026-08-04T02:07:14.094928+00:00"},{"alias_kind":"pith_short_12","alias_value":"E2P7QHWXKHA4","created_at":"2026-08-04T02:07:14.094928+00:00"},{"alias_kind":"pith_short_16","alias_value":"E2P7QHWXKHA4LDWO","created_at":"2026-08-04T02:07:14.094928+00:00"},{"alias_kind":"pith_short_8","alias_value":"E2P7QHWX","created_at":"2026-08-04T02:07:14.094928+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/E2P7QHWXKHA4LDWOAPATUSOOE4","json":"https://pith.science/pith/E2P7QHWXKHA4LDWOAPATUSOOE4.json","graph_json":"https://pith.science/api/pith-number/E2P7QHWXKHA4LDWOAPATUSOOE4/graph.json","events_json":"https://pith.science/api/pith-number/E2P7QHWXKHA4LDWOAPATUSOOE4/events.json","paper":"https://pith.science/paper/E2P7QHWX"},"agent_actions":{"view_html":"https://pith.science/pith/E2P7QHWXKHA4LDWOAPATUSOOE4","download_json":"https://pith.science/pith/E2P7QHWXKHA4LDWOAPATUSOOE4.json","view_paper":"https://pith.science/paper/E2P7QHWX","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2608.01816&json=true","fetch_graph":"https://pith.science/api/pith-number/E2P7QHWXKHA4LDWOAPATUSOOE4/graph.json","fetch_events":"https://pith.science/api/pith-number/E2P7QHWXKHA4LDWOAPATUSOOE4/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/E2P7QHWXKHA4LDWOAPATUSOOE4/action/timestamp_anchor","attest_storage":"https://pith.science/pith/E2P7QHWXKHA4LDWOAPATUSOOE4/action/storage_attestation","attest_author":"https://pith.science/pith/E2P7QHWXKHA4LDWOAPATUSOOE4/action/author_attestation","sign_citation":"https://pith.science/pith/E2P7QHWXKHA4LDWOAPATUSOOE4/action/citation_signature","submit_replication":"https://pith.science/pith/E2P7QHWXKHA4LDWOAPATUSOOE4/action/replication_record"}},"created_at":"2026-08-04T02:07:14.094928+00:00","updated_at":"2026-08-04T02:07:14.094928+00:00"}