{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:QHMYZJXO3NAMGEUQO7TAG5KTK3","short_pith_number":"pith:QHMYZJXO","schema_version":"1.0","canonical_sha256":"81d98ca6eedb40c3129077e603755356fd0f3fe8a5b79b8189569d1b651d278c","source":{"kind":"arxiv","id":"2601.04465","version":1},"attestation_state":"computed","paper":{"title":"Concept Tokens: Learning Behavioral Embeddings Through Concept Definitions","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.LG"],"primary_cat":"cs.CL","authors_text":"Aiala Ros\\'a, Ignacio Sastre","submitted_at":"2026-01-08T00:49:48Z","abstract_excerpt":"We propose Concept Tokens, a lightweight method that adds a new special token to a pretrained LLM and learns only its embedding from multiple natural language definitions of a target concept, where occurrences of the concept are replaced by the new token. The LLM is kept frozen and the embedding is optimized with the standard language-modeling objective. We evaluate Concept Tokens in three settings. First, we study hallucinations in closed-book question answering on HotpotQA and find a directional effect: negating the hallucination token reduces hallucinated answers mainly by increasing absten"},"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":"2601.04465","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2026-01-08T00:49:48Z","cross_cats_sorted":["cs.AI","cs.LG"],"title_canon_sha256":"9b833fcf76c7791e820da0e5a57c55b40b3b330e16225dde29a9471489da86f1","abstract_canon_sha256":"64b80f0a15848debae4d19de74e948abe822c2aa403f2b88029882909f18693d"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-29T01:25:33.016140Z","signature_b64":"LbclvgaGAMAFmkZ4fxi5zJaau8cryTkKHlGr/KEo5mP/dfDm/erzlTOrmo4bLFY+pxdTGOeSLrfjCL02RHTYAg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"81d98ca6eedb40c3129077e603755356fd0f3fe8a5b79b8189569d1b651d278c","last_reissued_at":"2026-07-29T01:25:33.015182Z","signature_status":"signed_v1","first_computed_at":"2026-07-29T01:25:33.015182Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Concept Tokens: Learning Behavioral Embeddings Through Concept Definitions","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.LG"],"primary_cat":"cs.CL","authors_text":"Aiala Ros\\'a, Ignacio Sastre","submitted_at":"2026-01-08T00:49:48Z","abstract_excerpt":"We propose Concept Tokens, a lightweight method that adds a new special token to a pretrained LLM and learns only its embedding from multiple natural language definitions of a target concept, where occurrences of the concept are replaced by the new token. The LLM is kept frozen and the embedding is optimized with the standard language-modeling objective. We evaluate Concept Tokens in three settings. First, we study hallucinations in closed-book question answering on HotpotQA and find a directional effect: negating the hallucination token reduces hallucinated answers mainly by increasing absten"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2601.04465","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/2601.04465/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":"2601.04465","created_at":"2026-07-29T01:25:33.015634+00:00"},{"alias_kind":"arxiv_version","alias_value":"2601.04465v1","created_at":"2026-07-29T01:25:33.015634+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2601.04465","created_at":"2026-07-29T01:25:33.015634+00:00"},{"alias_kind":"pith_short_12","alias_value":"QHMYZJXO3NAM","created_at":"2026-07-29T01:25:33.015634+00:00"},{"alias_kind":"pith_short_16","alias_value":"QHMYZJXO3NAMGEUQ","created_at":"2026-07-29T01:25:33.015634+00:00"},{"alias_kind":"pith_short_8","alias_value":"QHMYZJXO","created_at":"2026-07-29T01:25:33.015634+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/QHMYZJXO3NAMGEUQO7TAG5KTK3","json":"https://pith.science/pith/QHMYZJXO3NAMGEUQO7TAG5KTK3.json","graph_json":"https://pith.science/api/pith-number/QHMYZJXO3NAMGEUQO7TAG5KTK3/graph.json","events_json":"https://pith.science/api/pith-number/QHMYZJXO3NAMGEUQO7TAG5KTK3/events.json","paper":"https://pith.science/paper/QHMYZJXO"},"agent_actions":{"view_html":"https://pith.science/pith/QHMYZJXO3NAMGEUQO7TAG5KTK3","download_json":"https://pith.science/pith/QHMYZJXO3NAMGEUQO7TAG5KTK3.json","view_paper":"https://pith.science/paper/QHMYZJXO","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2601.04465&json=true","fetch_graph":"https://pith.science/api/pith-number/QHMYZJXO3NAMGEUQO7TAG5KTK3/graph.json","fetch_events":"https://pith.science/api/pith-number/QHMYZJXO3NAMGEUQO7TAG5KTK3/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/QHMYZJXO3NAMGEUQO7TAG5KTK3/action/timestamp_anchor","attest_storage":"https://pith.science/pith/QHMYZJXO3NAMGEUQO7TAG5KTK3/action/storage_attestation","attest_author":"https://pith.science/pith/QHMYZJXO3NAMGEUQO7TAG5KTK3/action/author_attestation","sign_citation":"https://pith.science/pith/QHMYZJXO3NAMGEUQO7TAG5KTK3/action/citation_signature","submit_replication":"https://pith.science/pith/QHMYZJXO3NAMGEUQO7TAG5KTK3/action/replication_record"}},"created_at":"2026-07-29T01:25:33.015634+00:00","updated_at":"2026-07-29T01:25:33.015634+00:00"}