{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:DSHZJTEF55OL3O7Z532GVD5HSN","short_pith_number":"pith:DSHZJTEF","schema_version":"1.0","canonical_sha256":"1c8f94cc85ef5cbdbbf9eef46a8fa79370dbcb35d1d5e06148b6832efabb1aeb","source":{"kind":"arxiv","id":"2606.00544","version":1},"attestation_state":"computed","paper":{"title":"Escaping the Mode Lottery: Multi-Response Training Improves Language Model Generalization","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.CL"],"primary_cat":"cs.LG","authors_text":"Hasan Amin, Kian Ahrabian, Ming Yin, Rajiv Khanna","submitted_at":"2026-05-30T05:30:39Z","abstract_excerpt":"Modern language-model fine-tuning typically pairs each prompt with a single response, even though many prompts admit multiple valid completions. This effectively reduces a multi-modal conditional distribution to a one-sample view, a phenomenon we call the \"mode lottery,\" where training emphasizes a subset of plausible modes while leaving others underrepresented. We study multi-response training (MRT), which retains multiple responses per prompt, and develop a principled account of when and why it helps. Our key insight is that prompts and responses are distinct statistical resources: additiona"},"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":"2606.00544","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-05-30T05:30:39Z","cross_cats_sorted":["cs.CL"],"title_canon_sha256":"ee866484c0414d8a91c6614b470b1692e9dd4bee400b057d825b5328cdcd5680","abstract_canon_sha256":"edabdb619954f2be68c1caebf0fe6a3ecf05f8f23e734b9c63cd13c1436403a7"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-06-02T01:03:57.766063Z","signature_b64":"84R7VuAYjdbMjEOCp09zuPQEaivT7BU+zq7EZKbaXUxX8rqIKzaSqobpdHFAGpX+4d8S2AdnYX9WTNHHuUBmDg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"1c8f94cc85ef5cbdbbf9eef46a8fa79370dbcb35d1d5e06148b6832efabb1aeb","last_reissued_at":"2026-06-02T01:03:57.765645Z","signature_status":"signed_v1","first_computed_at":"2026-06-02T01:03:57.765645Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Escaping the Mode Lottery: Multi-Response Training Improves Language Model Generalization","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.CL"],"primary_cat":"cs.LG","authors_text":"Hasan Amin, Kian Ahrabian, Ming Yin, Rajiv Khanna","submitted_at":"2026-05-30T05:30:39Z","abstract_excerpt":"Modern language-model fine-tuning typically pairs each prompt with a single response, even though many prompts admit multiple valid completions. This effectively reduces a multi-modal conditional distribution to a one-sample view, a phenomenon we call the \"mode lottery,\" where training emphasizes a subset of plausible modes while leaving others underrepresented. We study multi-response training (MRT), which retains multiple responses per prompt, and develop a principled account of when and why it helps. Our key insight is that prompts and responses are distinct statistical resources: additiona"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2606.00544","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/2606.00544/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":"2606.00544","created_at":"2026-06-02T01:03:57.765713+00:00"},{"alias_kind":"arxiv_version","alias_value":"2606.00544v1","created_at":"2026-06-02T01:03:57.765713+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2606.00544","created_at":"2026-06-02T01:03:57.765713+00:00"},{"alias_kind":"pith_short_12","alias_value":"DSHZJTEF55OL","created_at":"2026-06-02T01:03:57.765713+00:00"},{"alias_kind":"pith_short_16","alias_value":"DSHZJTEF55OL3O7Z","created_at":"2026-06-02T01:03:57.765713+00:00"},{"alias_kind":"pith_short_8","alias_value":"DSHZJTEF","created_at":"2026-06-02T01:03:57.765713+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/DSHZJTEF55OL3O7Z532GVD5HSN","json":"https://pith.science/pith/DSHZJTEF55OL3O7Z532GVD5HSN.json","graph_json":"https://pith.science/api/pith-number/DSHZJTEF55OL3O7Z532GVD5HSN/graph.json","events_json":"https://pith.science/api/pith-number/DSHZJTEF55OL3O7Z532GVD5HSN/events.json","paper":"https://pith.science/paper/DSHZJTEF"},"agent_actions":{"view_html":"https://pith.science/pith/DSHZJTEF55OL3O7Z532GVD5HSN","download_json":"https://pith.science/pith/DSHZJTEF55OL3O7Z532GVD5HSN.json","view_paper":"https://pith.science/paper/DSHZJTEF","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2606.00544&json=true","fetch_graph":"https://pith.science/api/pith-number/DSHZJTEF55OL3O7Z532GVD5HSN/graph.json","fetch_events":"https://pith.science/api/pith-number/DSHZJTEF55OL3O7Z532GVD5HSN/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/DSHZJTEF55OL3O7Z532GVD5HSN/action/timestamp_anchor","attest_storage":"https://pith.science/pith/DSHZJTEF55OL3O7Z532GVD5HSN/action/storage_attestation","attest_author":"https://pith.science/pith/DSHZJTEF55OL3O7Z532GVD5HSN/action/author_attestation","sign_citation":"https://pith.science/pith/DSHZJTEF55OL3O7Z532GVD5HSN/action/citation_signature","submit_replication":"https://pith.science/pith/DSHZJTEF55OL3O7Z532GVD5HSN/action/replication_record"}},"created_at":"2026-06-02T01:03:57.765713+00:00","updated_at":"2026-06-02T01:03:57.765713+00:00"}