{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2023:P5DB4HPUL56DYGXUGS7A677EIO","short_pith_number":"pith:P5DB4HPU","schema_version":"1.0","canonical_sha256":"7f461e1df45f7c3c1af434be0f7fe443be791615d8f99981f9ca50e71c964d72","source":{"kind":"arxiv","id":"2305.04990","version":3},"attestation_state":"computed","paper":{"title":"Explanation-based Finetuning Makes Models More Robust to Spurious Cues","license":"http://creativecommons.org/licenses/by-nc-nd/4.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"cs.CL","authors_text":"Chris Callison-Burch, Josh Magnus Ludan, Marianna Apidianaki, Qing Lyu, Saurabh Shah, Tai Nguyen, Yixuan Meng","submitted_at":"2023-05-08T18:53:45Z","abstract_excerpt":"Large Language Models (LLMs) are so powerful that they sometimes learn correlations between labels and features that are irrelevant to the task, leading to poor generalization on out-of-distribution data. We propose explanation-based finetuning as a general approach to mitigate LLMs' reliance on spurious correlations. Unlike standard finetuning where the model only predicts the answer given the input, we finetune the model to additionally generate a free-text explanation supporting its answer. To evaluate our method, we finetune the model on artificially constructed training sets containing di"},"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":"2305.04990","kind":"arxiv","version":3},"metadata":{"license":"http://creativecommons.org/licenses/by-nc-nd/4.0/","primary_cat":"cs.CL","submitted_at":"2023-05-08T18:53:45Z","cross_cats_sorted":["cs.LG"],"title_canon_sha256":"b3d7d828aee3450c7cd001b69e82394cf874fcaf46f1650d3ed557cd02fb2d3f","abstract_canon_sha256":"8903d2b8b109c561ed9bbc8ff6ff0a7890be6d084656bc40a2cac40d198e6729"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T06:17:51.083614Z","signature_b64":"RRcljqwo6sYJu5GY+vZZBa+rpgwapWijbOQL4hn54f0ys299H2RX/2u9JsvzYWP6IqAAMEFWwCngy4+CLxr/AA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"7f461e1df45f7c3c1af434be0f7fe443be791615d8f99981f9ca50e71c964d72","last_reissued_at":"2026-07-05T06:17:51.083128Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T06:17:51.083128Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Explanation-based Finetuning Makes Models More Robust to Spurious Cues","license":"http://creativecommons.org/licenses/by-nc-nd/4.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"cs.CL","authors_text":"Chris Callison-Burch, Josh Magnus Ludan, Marianna Apidianaki, Qing Lyu, Saurabh Shah, Tai Nguyen, Yixuan Meng","submitted_at":"2023-05-08T18:53:45Z","abstract_excerpt":"Large Language Models (LLMs) are so powerful that they sometimes learn correlations between labels and features that are irrelevant to the task, leading to poor generalization on out-of-distribution data. We propose explanation-based finetuning as a general approach to mitigate LLMs' reliance on spurious correlations. Unlike standard finetuning where the model only predicts the answer given the input, we finetune the model to additionally generate a free-text explanation supporting its answer. To evaluate our method, we finetune the model on artificially constructed training sets containing di"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2305.04990","kind":"arxiv","version":3},"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/2305.04990/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":"2305.04990","created_at":"2026-07-05T06:17:51.083194+00:00"},{"alias_kind":"arxiv_version","alias_value":"2305.04990v3","created_at":"2026-07-05T06:17:51.083194+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2305.04990","created_at":"2026-07-05T06:17:51.083194+00:00"},{"alias_kind":"pith_short_12","alias_value":"P5DB4HPUL56D","created_at":"2026-07-05T06:17:51.083194+00:00"},{"alias_kind":"pith_short_16","alias_value":"P5DB4HPUL56DYGXU","created_at":"2026-07-05T06:17:51.083194+00:00"},{"alias_kind":"pith_short_8","alias_value":"P5DB4HPU","created_at":"2026-07-05T06:17:51.083194+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2504.13898","citing_title":"Social Human Robot Embodied Conversation (SHREC) Dataset: Benchmarking Foundational Models' Social Reasoning","ref_index":35,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/P5DB4HPUL56DYGXUGS7A677EIO","json":"https://pith.science/pith/P5DB4HPUL56DYGXUGS7A677EIO.json","graph_json":"https://pith.science/api/pith-number/P5DB4HPUL56DYGXUGS7A677EIO/graph.json","events_json":"https://pith.science/api/pith-number/P5DB4HPUL56DYGXUGS7A677EIO/events.json","paper":"https://pith.science/paper/P5DB4HPU"},"agent_actions":{"view_html":"https://pith.science/pith/P5DB4HPUL56DYGXUGS7A677EIO","download_json":"https://pith.science/pith/P5DB4HPUL56DYGXUGS7A677EIO.json","view_paper":"https://pith.science/paper/P5DB4HPU","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2305.04990&json=true","fetch_graph":"https://pith.science/api/pith-number/P5DB4HPUL56DYGXUGS7A677EIO/graph.json","fetch_events":"https://pith.science/api/pith-number/P5DB4HPUL56DYGXUGS7A677EIO/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/P5DB4HPUL56DYGXUGS7A677EIO/action/timestamp_anchor","attest_storage":"https://pith.science/pith/P5DB4HPUL56DYGXUGS7A677EIO/action/storage_attestation","attest_author":"https://pith.science/pith/P5DB4HPUL56DYGXUGS7A677EIO/action/author_attestation","sign_citation":"https://pith.science/pith/P5DB4HPUL56DYGXUGS7A677EIO/action/citation_signature","submit_replication":"https://pith.science/pith/P5DB4HPUL56DYGXUGS7A677EIO/action/replication_record"}},"created_at":"2026-07-05T06:17:51.083194+00:00","updated_at":"2026-07-05T06:17:51.083194+00:00"}