{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:2X4G4LALNI46YHEY7XNRPKORQG","short_pith_number":"pith:2X4G4LAL","schema_version":"1.0","canonical_sha256":"d5f86e2c0b6a39ec1c98fddb17a9d181be75692972af1c7ee1f95c9028732f92","source":{"kind":"arxiv","id":"2508.07616","version":2},"attestation_state":"computed","paper":{"title":"ThinkTuning: Instilling Cognitive Reflections without Distillation","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.CL","cs.LG"],"primary_cat":"cs.AI","authors_text":"Aswin RRV, Ben Zhou, Chitta Baral, Divij Handa, Jacob Dineen, Md Nayem Uddin, Mihir Parmar","submitted_at":"2025-08-11T04:51:43Z","abstract_excerpt":"Recent advances in test-time scaling have led to the emergence of thinking LLMs that exhibit self-reflective behaviors and multi-step reasoning. While RL drives this self-improvement paradigm, a recent study (Gandhi et al., 2025) shows that RL alone does not truly instill these new reasoning abilities - it merely draws out behaviors already present in the base models. This raises a question: How can we train the models that don't exhibit such thinking behavior to develop it in the first place? To this end, we propose ThinkTuning, a GRPO-based interactive training approach where we augment the "},"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":"2508.07616","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.AI","submitted_at":"2025-08-11T04:51:43Z","cross_cats_sorted":["cs.CL","cs.LG"],"title_canon_sha256":"57d2a215a8e77ead83a8fa12920f43783414abae8ab09c14215482a5a5e3054b","abstract_canon_sha256":"7f920255c97b215b8c8a72d49d9661b3f7bcf369944c62595c29f4ea2ee7ab24"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:57:00.971174Z","signature_b64":"s72p6Gm2Uzngg+f57dT3U8BXWNLL+9HWFE81Zd48al3Q/txiVRG30uxAeuqVEUN34grZdyPlDIDW/jo2NN+dDA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"d5f86e2c0b6a39ec1c98fddb17a9d181be75692972af1c7ee1f95c9028732f92","last_reissued_at":"2026-07-05T11:57:00.970717Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:57:00.970717Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"ThinkTuning: Instilling Cognitive Reflections without Distillation","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.CL","cs.LG"],"primary_cat":"cs.AI","authors_text":"Aswin RRV, Ben Zhou, Chitta Baral, Divij Handa, Jacob Dineen, Md Nayem Uddin, Mihir Parmar","submitted_at":"2025-08-11T04:51:43Z","abstract_excerpt":"Recent advances in test-time scaling have led to the emergence of thinking LLMs that exhibit self-reflective behaviors and multi-step reasoning. While RL drives this self-improvement paradigm, a recent study (Gandhi et al., 2025) shows that RL alone does not truly instill these new reasoning abilities - it merely draws out behaviors already present in the base models. This raises a question: How can we train the models that don't exhibit such thinking behavior to develop it in the first place? To this end, we propose ThinkTuning, a GRPO-based interactive training approach where we augment the "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2508.07616","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/2508.07616/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":"2508.07616","created_at":"2026-07-05T11:57:00.970775+00:00"},{"alias_kind":"arxiv_version","alias_value":"2508.07616v2","created_at":"2026-07-05T11:57:00.970775+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2508.07616","created_at":"2026-07-05T11:57:00.970775+00:00"},{"alias_kind":"pith_short_12","alias_value":"2X4G4LALNI46","created_at":"2026-07-05T11:57:00.970775+00:00"},{"alias_kind":"pith_short_16","alias_value":"2X4G4LALNI46YHEY","created_at":"2026-07-05T11:57:00.970775+00:00"},{"alias_kind":"pith_short_8","alias_value":"2X4G4LAL","created_at":"2026-07-05T11:57:00.970775+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":2,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2509.10746","citing_title":"RECAP: Transparent Inference-Time Emotion Alignment for Medical Dialogue Systems","ref_index":40,"is_internal_anchor":false},{"citing_arxiv_id":"2605.08472","citing_title":"Mid-Training with Self-Generated Data Improves Reinforcement Learning in Language Models","ref_index":42,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/2X4G4LALNI46YHEY7XNRPKORQG","json":"https://pith.science/pith/2X4G4LALNI46YHEY7XNRPKORQG.json","graph_json":"https://pith.science/api/pith-number/2X4G4LALNI46YHEY7XNRPKORQG/graph.json","events_json":"https://pith.science/api/pith-number/2X4G4LALNI46YHEY7XNRPKORQG/events.json","paper":"https://pith.science/paper/2X4G4LAL"},"agent_actions":{"view_html":"https://pith.science/pith/2X4G4LALNI46YHEY7XNRPKORQG","download_json":"https://pith.science/pith/2X4G4LALNI46YHEY7XNRPKORQG.json","view_paper":"https://pith.science/paper/2X4G4LAL","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2508.07616&json=true","fetch_graph":"https://pith.science/api/pith-number/2X4G4LALNI46YHEY7XNRPKORQG/graph.json","fetch_events":"https://pith.science/api/pith-number/2X4G4LALNI46YHEY7XNRPKORQG/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/2X4G4LALNI46YHEY7XNRPKORQG/action/timestamp_anchor","attest_storage":"https://pith.science/pith/2X4G4LALNI46YHEY7XNRPKORQG/action/storage_attestation","attest_author":"https://pith.science/pith/2X4G4LALNI46YHEY7XNRPKORQG/action/author_attestation","sign_citation":"https://pith.science/pith/2X4G4LALNI46YHEY7XNRPKORQG/action/citation_signature","submit_replication":"https://pith.science/pith/2X4G4LALNI46YHEY7XNRPKORQG/action/replication_record"}},"created_at":"2026-07-05T11:57:00.970775+00:00","updated_at":"2026-07-05T11:57:00.970775+00:00"}