{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:HWBWMP5YXRRWOQTVRXRUMCZTLK","short_pith_number":"pith:HWBWMP5Y","schema_version":"1.0","canonical_sha256":"3d83663fb8bc636742758de3460b335a93eb726f3a76521530b9bf4ec57bae6d","source":{"kind":"arxiv","id":"2605.13369","version":2},"attestation_state":"computed","paper":{"title":"Query-Conditioned Test-Time Self-Training for Large Language Models","license":"http://creativecommons.org/licenses/by/4.0/","headline":"Large language models can adapt their own parameters during inference by generating training examples directly from the input query.","cross_cats":["cs.AI","cs.LG"],"primary_cat":"cs.CL","authors_text":"Chaehee Song, Changick Kim, Doyi Kim, Minseok Seo, Yeeun Seong","submitted_at":"2026-05-13T11:27:40Z","abstract_excerpt":"Large language models (LLMs) are typically deployed with fixed parameters, and their performance is often improved by allocating more computation at inference time. While such test-time scaling can be effective, it cannot correct model misconceptions or adapt the model to the specific structure of an individual query. Test-time optimization addresses this limitation by enabling parameter updates during inference, but existing approaches either rely on external data or optimize generic self-supervised objectives that lack query-specific alignment. In this work, we propose Query-Conditioned Test"},"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":true,"formal_links_present":true},"canonical_record":{"source":{"id":"2605.13369","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2026-05-13T11:27:40Z","cross_cats_sorted":["cs.AI","cs.LG"],"title_canon_sha256":"129407ebd1cd3bcd2614d4223dc14cb773fdb0c6fff23a384f258abcd960d544","abstract_canon_sha256":"f8740c8dc59a74afea06d0eb91d97dd84ca9a0c9f9fcf3a24315c6ad7d783b4a"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-18T02:44:47.996582Z","signature_b64":"Q6Wwdeq1A8pB46V12doBw8fn7NvaVrkN3TCJzffcrf4+4uXrarF6uejmFShh2aO5l3TxTIyXrHjSKYpMFmYkAg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"3d83663fb8bc636742758de3460b335a93eb726f3a76521530b9bf4ec57bae6d","last_reissued_at":"2026-05-18T02:44:47.996156Z","signature_status":"signed_v1","first_computed_at":"2026-05-18T02:44:47.996156Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Query-Conditioned Test-Time Self-Training for Large Language Models","license":"http://creativecommons.org/licenses/by/4.0/","headline":"Large language models can adapt their own parameters during inference by generating training examples directly from the input query.","cross_cats":["cs.AI","cs.LG"],"primary_cat":"cs.CL","authors_text":"Chaehee Song, Changick Kim, Doyi Kim, Minseok Seo, Yeeun Seong","submitted_at":"2026-05-13T11:27:40Z","abstract_excerpt":"Large language models (LLMs) are typically deployed with fixed parameters, and their performance is often improved by allocating more computation at inference time. While such test-time scaling can be effective, it cannot correct model misconceptions or adapt the model to the specific structure of an individual query. Test-time optimization addresses this limitation by enabling parameter updates during inference, but existing approaches either rely on external data or optimize generic self-supervised objectives that lack query-specific alignment. In this work, we propose Query-Conditioned Test"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"QueST generates such query-conditioned pairs and uses them as supervision for parameter-efficient fine-tuning at test time. The adapted model is then used to produce the final answer, enabling query-specific adaptation without any external data.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"the input query itself encodes latent signals sufficient for constructing structurally related problem--solution pairs","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"QueST adapts LLMs at test time by generating query-specific problem-solution pairs for self-supervised fine-tuning, improving reasoning performance without external data.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"Large language models can adapt their own parameters during inference by generating training examples directly from the input query.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"b88bff5c21af8a3ba49945f1935cb529374619927cb482b0062bb67fce8e2df1"},"source":{"id":"2605.13369","kind":"arxiv","version":2},"verdict":{"id":"19d914b4-30f0-4f16-a522-95bf33f7836c","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-15T05:50:41.544249Z","strongest_claim":"QueST generates such query-conditioned pairs and uses them as supervision for parameter-efficient fine-tuning at test time. The adapted model is then used to produce the final answer, enabling query-specific adaptation without any external data.","one_line_summary":"QueST adapts LLMs at test time by generating query-specific problem-solution pairs for self-supervised fine-tuning, improving reasoning performance without external data.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"the input query itself encodes latent signals sufficient for constructing structurally related problem--solution pairs","pith_extraction_headline":"Large language models can adapt their own parameters during inference by generating training examples directly from the input query."},"references":{"count":42,"sample":[{"doi":"","year":2023,"title":"GPT-4 Technical Report","work_id":"b928e041-6991-4c08-8c81-0359e4097c7b","ref_index":1,"cited_arxiv_id":"2303.08774","is_internal_anchor":true},{"doi":"","year":2021,"title":"Training Verifiers to Solve Math Word Problems","work_id":"acab1aa8-b4d6-40e0-a3ee-25341701dca2","ref_index":2,"cited_arxiv_id":"2110.14168","is_internal_anchor":true},{"doi":"","year":2026,"title":"In-Place Test-Time Training","work_id":"7edff7e1-6263-47c6-afd5-e77029c3948a","ref_index":3,"cited_arxiv_id":"2604.06169","is_internal_anchor":true},{"doi":"","year":2025,"title":"DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learning","work_id":"e6b75ad5-2877-4168-97c8-710407094d20","ref_index":4,"cited_arxiv_id":"2501.12948","is_internal_anchor":true},{"doi":"","year":2023,"title":"Test-time training on nearest neighbors for large language models.arXiv preprint arXiv:2305.18466","work_id":"88670479-3b0a-4a44-8a0d-d11feb3aefdd","ref_index":5,"cited_arxiv_id":"","is_internal_anchor":false}],"resolved_work":42,"snapshot_sha256":"7c533a3b618ccf95b732bc4a5aae4b75fba986ea15a8346514bd0429d8cddd2b","internal_anchors":14},"formal_canon":{"evidence_count":2,"snapshot_sha256":"f3b93275f7df4723275dca8b4d28e92e0e092133e3be7eff10726c1eb13d2568"},"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":"2605.13369","created_at":"2026-05-18T02:44:47.996225+00:00"},{"alias_kind":"arxiv_version","alias_value":"2605.13369v2","created_at":"2026-05-18T02:44:47.996225+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.13369","created_at":"2026-05-18T02:44:47.996225+00:00"},{"alias_kind":"pith_short_12","alias_value":"HWBWMP5YXRRW","created_at":"2026-05-18T12:33:37.589309+00:00"},{"alias_kind":"pith_short_16","alias_value":"HWBWMP5YXRRWOQTV","created_at":"2026-05-18T12:33:37.589309+00:00"},{"alias_kind":"pith_short_8","alias_value":"HWBWMP5Y","created_at":"2026-05-18T12:33:37.589309+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":0,"internal_anchor_count":0,"sample":[]},"formal_canon":{"evidence_count":2,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/HWBWMP5YXRRWOQTVRXRUMCZTLK","json":"https://pith.science/pith/HWBWMP5YXRRWOQTVRXRUMCZTLK.json","graph_json":"https://pith.science/api/pith-number/HWBWMP5YXRRWOQTVRXRUMCZTLK/graph.json","events_json":"https://pith.science/api/pith-number/HWBWMP5YXRRWOQTVRXRUMCZTLK/events.json","paper":"https://pith.science/paper/HWBWMP5Y"},"agent_actions":{"view_html":"https://pith.science/pith/HWBWMP5YXRRWOQTVRXRUMCZTLK","download_json":"https://pith.science/pith/HWBWMP5YXRRWOQTVRXRUMCZTLK.json","view_paper":"https://pith.science/paper/HWBWMP5Y","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2605.13369&json=true","fetch_graph":"https://pith.science/api/pith-number/HWBWMP5YXRRWOQTVRXRUMCZTLK/graph.json","fetch_events":"https://pith.science/api/pith-number/HWBWMP5YXRRWOQTVRXRUMCZTLK/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/HWBWMP5YXRRWOQTVRXRUMCZTLK/action/timestamp_anchor","attest_storage":"https://pith.science/pith/HWBWMP5YXRRWOQTVRXRUMCZTLK/action/storage_attestation","attest_author":"https://pith.science/pith/HWBWMP5YXRRWOQTVRXRUMCZTLK/action/author_attestation","sign_citation":"https://pith.science/pith/HWBWMP5YXRRWOQTVRXRUMCZTLK/action/citation_signature","submit_replication":"https://pith.science/pith/HWBWMP5YXRRWOQTVRXRUMCZTLK/action/replication_record"}},"created_at":"2026-05-18T02:44:47.996225+00:00","updated_at":"2026-05-18T02:44:47.996225+00:00"}