{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:AQW4SOVVZ2K3DINNFT2KDMB4Q2","short_pith_number":"pith:AQW4SOVV","schema_version":"1.0","canonical_sha256":"042dc93ab5ce95b1a1ad2cf4a1b03c869de0019b640575cc8f95e37d72e3c5e0","source":{"kind":"arxiv","id":"2408.00113","version":2},"attestation_state":"computed","paper":{"title":"Measuring Progress in Dictionary Learning for Language Model Interpretability with Board Game Models","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.CL"],"primary_cat":"cs.LG","authors_text":"Adam Karvonen, Benjamin Wright, Can Rager, Claudio Mayrink Verdun, David Bau, Jannik Brinkmann, Logan Smith, Rico Angell, Samuel Marks","submitted_at":"2024-07-31T18:45:13Z","abstract_excerpt":"What latent features are encoded in language model (LM) representations? Recent work on training sparse autoencoders (SAEs) to disentangle interpretable features in LM representations has shown significant promise. However, evaluating the quality of these SAEs is difficult because we lack a ground-truth collection of interpretable features that we expect good SAEs to recover. We thus propose to measure progress in interpretable dictionary learning by working in the setting of LMs trained on chess and Othello transcripts. These settings carry natural collections of interpretable features -- for"},"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":"2408.00113","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2024-07-31T18:45:13Z","cross_cats_sorted":["cs.AI","cs.CL"],"title_canon_sha256":"aed5adc4132b59ae7381121edba1fca1aeb7b19168f365b7bf9ae2326b83a559","abstract_canon_sha256":"a36fec959115f82fba90ae4cd82e5f1176e367d7ef2cb2caa9b6f151e72f995b"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T09:28:33.650856Z","signature_b64":"iMfViB+KalUoIxj7foI+kOFMf9IWL9dj1NxCmer5uq71SD5uGwX3hs8LgaxU5CrG0GnZPl5tZSg/TxtRQeNIDQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"042dc93ab5ce95b1a1ad2cf4a1b03c869de0019b640575cc8f95e37d72e3c5e0","last_reissued_at":"2026-07-05T09:28:33.650361Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T09:28:33.650361Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Measuring Progress in Dictionary Learning for Language Model Interpretability with Board Game Models","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.CL"],"primary_cat":"cs.LG","authors_text":"Adam Karvonen, Benjamin Wright, Can Rager, Claudio Mayrink Verdun, David Bau, Jannik Brinkmann, Logan Smith, Rico Angell, Samuel Marks","submitted_at":"2024-07-31T18:45:13Z","abstract_excerpt":"What latent features are encoded in language model (LM) representations? Recent work on training sparse autoencoders (SAEs) to disentangle interpretable features in LM representations has shown significant promise. However, evaluating the quality of these SAEs is difficult because we lack a ground-truth collection of interpretable features that we expect good SAEs to recover. We thus propose to measure progress in interpretable dictionary learning by working in the setting of LMs trained on chess and Othello transcripts. These settings carry natural collections of interpretable features -- for"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2408.00113","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/2408.00113/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":"2408.00113","created_at":"2026-07-05T09:28:33.650416+00:00"},{"alias_kind":"arxiv_version","alias_value":"2408.00113v2","created_at":"2026-07-05T09:28:33.650416+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2408.00113","created_at":"2026-07-05T09:28:33.650416+00:00"},{"alias_kind":"pith_short_12","alias_value":"AQW4SOVVZ2K3","created_at":"2026-07-05T09:28:33.650416+00:00"},{"alias_kind":"pith_short_16","alias_value":"AQW4SOVVZ2K3DINN","created_at":"2026-07-05T09:28:33.650416+00:00"},{"alias_kind":"pith_short_8","alias_value":"AQW4SOVV","created_at":"2026-07-05T09:28:33.650416+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":2,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2605.28149","citing_title":"Sign-Aware Gated Sparse Autoencoders: Modeling Anticorrelated Features with Bi-Jump-ReLU Activations","ref_index":43,"is_internal_anchor":false},{"citing_arxiv_id":"2509.18127","citing_title":"Safe-SAIL: Towards a Fine-grained Safety Landscape of Large Language Models via Sparse Autoencoder Interpretation Framework","ref_index":21,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/AQW4SOVVZ2K3DINNFT2KDMB4Q2","json":"https://pith.science/pith/AQW4SOVVZ2K3DINNFT2KDMB4Q2.json","graph_json":"https://pith.science/api/pith-number/AQW4SOVVZ2K3DINNFT2KDMB4Q2/graph.json","events_json":"https://pith.science/api/pith-number/AQW4SOVVZ2K3DINNFT2KDMB4Q2/events.json","paper":"https://pith.science/paper/AQW4SOVV"},"agent_actions":{"view_html":"https://pith.science/pith/AQW4SOVVZ2K3DINNFT2KDMB4Q2","download_json":"https://pith.science/pith/AQW4SOVVZ2K3DINNFT2KDMB4Q2.json","view_paper":"https://pith.science/paper/AQW4SOVV","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2408.00113&json=true","fetch_graph":"https://pith.science/api/pith-number/AQW4SOVVZ2K3DINNFT2KDMB4Q2/graph.json","fetch_events":"https://pith.science/api/pith-number/AQW4SOVVZ2K3DINNFT2KDMB4Q2/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/AQW4SOVVZ2K3DINNFT2KDMB4Q2/action/timestamp_anchor","attest_storage":"https://pith.science/pith/AQW4SOVVZ2K3DINNFT2KDMB4Q2/action/storage_attestation","attest_author":"https://pith.science/pith/AQW4SOVVZ2K3DINNFT2KDMB4Q2/action/author_attestation","sign_citation":"https://pith.science/pith/AQW4SOVVZ2K3DINNFT2KDMB4Q2/action/citation_signature","submit_replication":"https://pith.science/pith/AQW4SOVVZ2K3DINNFT2KDMB4Q2/action/replication_record"}},"created_at":"2026-07-05T09:28:33.650416+00:00","updated_at":"2026-07-05T09:28:33.650416+00:00"}