{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:E63HPCP6EI6MFODASROWKI3X6E","short_pith_number":"pith:E63HPCP6","schema_version":"1.0","canonical_sha256":"27b67789fe223cc2b860945d652377f11ec5e8c88d706d014333900bcda68419","source":{"kind":"arxiv","id":"2502.01615","version":2},"attestation_state":"computed","paper":{"title":"Large Language Models Are Human-Like Internally","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Kentaro Inui, Souhaib Ben Taieb, Tatsuki Kuribayashi, Timothy Baldwin, Yohei Oseki","submitted_at":"2025-02-03T18:48:32Z","abstract_excerpt":"Recent cognitive modeling studies have reported that larger language models (LMs) exhibit a poorer fit to human reading behavior (Oh and Schuler, 2023b; Shain et al., 2024; Kuribayashi et al., 2024), leading to claims of their cognitive implausibility. In this paper, we revisit this argument through the lens of mechanistic interpretability and argue that prior conclusions were skewed by an exclusive focus on the final layers of LMs. Our analysis reveals that next-word probabilities derived from internal layers of larger LMs align with human sentence processing data as well as, or better than, "},"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":"2502.01615","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2025-02-03T18:48:32Z","cross_cats_sorted":[],"title_canon_sha256":"8710bf0007d69ebdbc272ad6663e97268305435c9f3f1cca8e307add44a89e31","abstract_canon_sha256":"36a528911111386b19c2efde4e3325d07dd80bc53e3f811c03066d198e20834e"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:43:30.543651Z","signature_b64":"aW0j0u/3WMGdzqq+++FZgJTnh4HbY3jaDWaTQGDlX5ylIFKRSGF6gDeVTQXQh5w3/k338MUIbFjZx0agsFhYCw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"27b67789fe223cc2b860945d652377f11ec5e8c88d706d014333900bcda68419","last_reissued_at":"2026-07-05T11:43:30.543153Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:43:30.543153Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Large Language Models Are Human-Like Internally","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Kentaro Inui, Souhaib Ben Taieb, Tatsuki Kuribayashi, Timothy Baldwin, Yohei Oseki","submitted_at":"2025-02-03T18:48:32Z","abstract_excerpt":"Recent cognitive modeling studies have reported that larger language models (LMs) exhibit a poorer fit to human reading behavior (Oh and Schuler, 2023b; Shain et al., 2024; Kuribayashi et al., 2024), leading to claims of their cognitive implausibility. In this paper, we revisit this argument through the lens of mechanistic interpretability and argue that prior conclusions were skewed by an exclusive focus on the final layers of LMs. Our analysis reveals that next-word probabilities derived from internal layers of larger LMs align with human sentence processing data as well as, or better than, "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2502.01615","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/2502.01615/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":"2502.01615","created_at":"2026-07-05T11:43:30.543211+00:00"},{"alias_kind":"arxiv_version","alias_value":"2502.01615v2","created_at":"2026-07-05T11:43:30.543211+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2502.01615","created_at":"2026-07-05T11:43:30.543211+00:00"},{"alias_kind":"pith_short_12","alias_value":"E63HPCP6EI6M","created_at":"2026-07-05T11:43:30.543211+00:00"},{"alias_kind":"pith_short_16","alias_value":"E63HPCP6EI6MFODA","created_at":"2026-07-05T11:43:30.543211+00:00"},{"alias_kind":"pith_short_8","alias_value":"E63HPCP6","created_at":"2026-07-05T11:43:30.543211+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":3,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.01671","citing_title":"When Meaning Travels: A Granular Lens on Hybrid-MoE's Role in Idiomatic Understanding for Language Models","ref_index":105,"is_internal_anchor":false},{"citing_arxiv_id":"2606.29904","citing_title":"Timesteps of Mamba Align with Human Reading Times","ref_index":16,"is_internal_anchor":false},{"citing_arxiv_id":"2605.15440","citing_title":"Why are language models less surprised than humans? Testing the Parse Multiplicity Mismatch Hypothesis","ref_index":166,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/E63HPCP6EI6MFODASROWKI3X6E","json":"https://pith.science/pith/E63HPCP6EI6MFODASROWKI3X6E.json","graph_json":"https://pith.science/api/pith-number/E63HPCP6EI6MFODASROWKI3X6E/graph.json","events_json":"https://pith.science/api/pith-number/E63HPCP6EI6MFODASROWKI3X6E/events.json","paper":"https://pith.science/paper/E63HPCP6"},"agent_actions":{"view_html":"https://pith.science/pith/E63HPCP6EI6MFODASROWKI3X6E","download_json":"https://pith.science/pith/E63HPCP6EI6MFODASROWKI3X6E.json","view_paper":"https://pith.science/paper/E63HPCP6","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2502.01615&json=true","fetch_graph":"https://pith.science/api/pith-number/E63HPCP6EI6MFODASROWKI3X6E/graph.json","fetch_events":"https://pith.science/api/pith-number/E63HPCP6EI6MFODASROWKI3X6E/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/E63HPCP6EI6MFODASROWKI3X6E/action/timestamp_anchor","attest_storage":"https://pith.science/pith/E63HPCP6EI6MFODASROWKI3X6E/action/storage_attestation","attest_author":"https://pith.science/pith/E63HPCP6EI6MFODASROWKI3X6E/action/author_attestation","sign_citation":"https://pith.science/pith/E63HPCP6EI6MFODASROWKI3X6E/action/citation_signature","submit_replication":"https://pith.science/pith/E63HPCP6EI6MFODASROWKI3X6E/action/replication_record"}},"created_at":"2026-07-05T11:43:30.543211+00:00","updated_at":"2026-07-05T11:43:30.543211+00:00"}