{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2023:WF622DSBZA26JOGTATXBNZOZZ4","short_pith_number":"pith:WF622DSB","schema_version":"1.0","canonical_sha256":"b17dad0e41c835e4b8d304ee16e5d9cf3b46dda726205d97d46e51dcee159ff7","source":{"kind":"arxiv","id":"2306.10062","version":1},"attestation_state":"computed","paper":{"title":"Revealing the structure of language model capabilities","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","cs.LG"],"primary_cat":"cs.CL","authors_text":"Andrew R. A. Conway, Han Hao, Jose Hernandez Orallo, Ryan Burnell","submitted_at":"2023-06-14T15:43:25Z","abstract_excerpt":"Building a theoretical understanding of the capabilities of large language models (LLMs) is vital for our ability to predict and explain the behavior of these systems. Here, we investigate the structure of LLM capabilities by extracting latent capabilities from patterns of individual differences across a varied population of LLMs. Using a combination of Bayesian and frequentist factor analysis, we analyzed data from 29 different LLMs across 27 cognitive tasks. We found evidence that LLM capabilities are not monolithic. Instead, they are better explained by three well-delineated factors that re"},"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":"2306.10062","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2023-06-14T15:43:25Z","cross_cats_sorted":["cs.AI","cs.LG"],"title_canon_sha256":"8f53767ada2281bb9d78f6d96f9e1e19364dfa664a98eb34d1421fb27ec06843","abstract_canon_sha256":"5790fe87d8ab19e8c928fe5695ece36c6f96fcd293948ee11873cfd67beab593"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T06:21:33.190885Z","signature_b64":"SbSVOlRrZgMWOwGi8xUA2zIOtZna6m3T0z9HrA7jnDAWV5+gz+8f/fJVXb+EkSM2U/B1slmZm9xx4ytVCePaBQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"b17dad0e41c835e4b8d304ee16e5d9cf3b46dda726205d97d46e51dcee159ff7","last_reissued_at":"2026-07-05T06:21:33.190409Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T06:21:33.190409Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Revealing the structure of language model capabilities","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","cs.LG"],"primary_cat":"cs.CL","authors_text":"Andrew R. A. Conway, Han Hao, Jose Hernandez Orallo, Ryan Burnell","submitted_at":"2023-06-14T15:43:25Z","abstract_excerpt":"Building a theoretical understanding of the capabilities of large language models (LLMs) is vital for our ability to predict and explain the behavior of these systems. Here, we investigate the structure of LLM capabilities by extracting latent capabilities from patterns of individual differences across a varied population of LLMs. Using a combination of Bayesian and frequentist factor analysis, we analyzed data from 29 different LLMs across 27 cognitive tasks. We found evidence that LLM capabilities are not monolithic. Instead, they are better explained by three well-delineated factors that re"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2306.10062","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/2306.10062/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":"2306.10062","created_at":"2026-07-05T06:21:33.190465+00:00"},{"alias_kind":"arxiv_version","alias_value":"2306.10062v1","created_at":"2026-07-05T06:21:33.190465+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2306.10062","created_at":"2026-07-05T06:21:33.190465+00:00"},{"alias_kind":"pith_short_12","alias_value":"WF622DSBZA26","created_at":"2026-07-05T06:21:33.190465+00:00"},{"alias_kind":"pith_short_16","alias_value":"WF622DSBZA26JOGT","created_at":"2026-07-05T06:21:33.190465+00:00"},{"alias_kind":"pith_short_8","alias_value":"WF622DSB","created_at":"2026-07-05T06:21:33.190465+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":6,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.24020","citing_title":"You Don't Need to Run Every Eval","ref_index":25,"is_internal_anchor":false},{"citing_arxiv_id":"2607.02464","citing_title":"Will Scaling Improve Social Simulation with LLMs?","ref_index":2,"is_internal_anchor":false},{"citing_arxiv_id":"2606.02981","citing_title":"Predicting Inference-Time Scaling Gains from Labeled Validation-Set Output Statistics","ref_index":3,"is_internal_anchor":false},{"citing_arxiv_id":"2605.24299","citing_title":"LLMs Show No Signs Of Individuated Metacognition","ref_index":4,"is_internal_anchor":false},{"citing_arxiv_id":"2602.18911","citing_title":"From Human-Level AI Tales to AI Leveling Human Scales","ref_index":2,"is_internal_anchor":false},{"citing_arxiv_id":"2604.12843","citing_title":"Growing Pains: Extensible and Efficient LLM Benchmarking Via Fixed Parameter Calibration","ref_index":1,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/WF622DSBZA26JOGTATXBNZOZZ4","json":"https://pith.science/pith/WF622DSBZA26JOGTATXBNZOZZ4.json","graph_json":"https://pith.science/api/pith-number/WF622DSBZA26JOGTATXBNZOZZ4/graph.json","events_json":"https://pith.science/api/pith-number/WF622DSBZA26JOGTATXBNZOZZ4/events.json","paper":"https://pith.science/paper/WF622DSB"},"agent_actions":{"view_html":"https://pith.science/pith/WF622DSBZA26JOGTATXBNZOZZ4","download_json":"https://pith.science/pith/WF622DSBZA26JOGTATXBNZOZZ4.json","view_paper":"https://pith.science/paper/WF622DSB","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2306.10062&json=true","fetch_graph":"https://pith.science/api/pith-number/WF622DSBZA26JOGTATXBNZOZZ4/graph.json","fetch_events":"https://pith.science/api/pith-number/WF622DSBZA26JOGTATXBNZOZZ4/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/WF622DSBZA26JOGTATXBNZOZZ4/action/timestamp_anchor","attest_storage":"https://pith.science/pith/WF622DSBZA26JOGTATXBNZOZZ4/action/storage_attestation","attest_author":"https://pith.science/pith/WF622DSBZA26JOGTATXBNZOZZ4/action/author_attestation","sign_citation":"https://pith.science/pith/WF622DSBZA26JOGTATXBNZOZZ4/action/citation_signature","submit_replication":"https://pith.science/pith/WF622DSBZA26JOGTATXBNZOZZ4/action/replication_record"}},"created_at":"2026-07-05T06:21:33.190465+00:00","updated_at":"2026-07-05T06:21:33.190465+00:00"}