{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:SDAS6GF2L724EKGUIWJ2LKZ7HX","short_pith_number":"pith:SDAS6GF2","schema_version":"1.0","canonical_sha256":"90c12f18ba5ff5c228d44593a5ab3f3deceefb04799e4791228106169a004b19","source":{"kind":"arxiv","id":"2505.19501","version":2},"attestation_state":"computed","paper":{"title":"Toward Scientific Reasoning in LLMs: Training from Expert Discussions via Reinforcement Learning","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.AI","authors_text":"Le Cong, Ling Yang, Mengdi Wang, Ming Yin, Yuanhao Qu","submitted_at":"2025-05-26T04:28:46Z","abstract_excerpt":"We investigate how to teach large language models (LLMs) to perform scientific reasoning by leveraging expert discussions as a learning signal. Focusing on the genomics domain, we develop an automated pipeline to extract trainable data and introduce Genome-Bench, a new benchmark constructed from over a decade of scientific forum discussions on genome engineering. Our pipeline transforms raw interactions into a reinforcement learning-friendly multiple-choice questions format, supported by 3000+ high-quality question-answer pairs spanning foundational biology, experimental troubleshooting, tool "},"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":"2505.19501","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.AI","submitted_at":"2025-05-26T04:28:46Z","cross_cats_sorted":[],"title_canon_sha256":"0fb067474eacedbab863573310682e5b330b79202f0dbb2f3877b3092ec9c761","abstract_canon_sha256":"e7fc3499fc0744f48009e5beae82b33e63ca44516373bbabe688f8017c4c7472"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:14:29.522712Z","signature_b64":"jgFBofvofDlrf64CXk8z8hC//MLQf3eSRgRhccGHDZ7NG+F4PUHBP7t+Gx5J6up6FNXOQaj3Ll8toUeplqNCCg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"90c12f18ba5ff5c228d44593a5ab3f3deceefb04799e4791228106169a004b19","last_reissued_at":"2026-07-05T11:14:29.522230Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:14:29.522230Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Toward Scientific Reasoning in LLMs: Training from Expert Discussions via Reinforcement Learning","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.AI","authors_text":"Le Cong, Ling Yang, Mengdi Wang, Ming Yin, Yuanhao Qu","submitted_at":"2025-05-26T04:28:46Z","abstract_excerpt":"We investigate how to teach large language models (LLMs) to perform scientific reasoning by leveraging expert discussions as a learning signal. Focusing on the genomics domain, we develop an automated pipeline to extract trainable data and introduce Genome-Bench, a new benchmark constructed from over a decade of scientific forum discussions on genome engineering. Our pipeline transforms raw interactions into a reinforcement learning-friendly multiple-choice questions format, supported by 3000+ high-quality question-answer pairs spanning foundational biology, experimental troubleshooting, tool "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2505.19501","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/2505.19501/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":"2505.19501","created_at":"2026-07-05T11:14:29.522287+00:00"},{"alias_kind":"arxiv_version","alias_value":"2505.19501v2","created_at":"2026-07-05T11:14:29.522287+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2505.19501","created_at":"2026-07-05T11:14:29.522287+00:00"},{"alias_kind":"pith_short_12","alias_value":"SDAS6GF2L724","created_at":"2026-07-05T11:14:29.522287+00:00"},{"alias_kind":"pith_short_16","alias_value":"SDAS6GF2L724EKGU","created_at":"2026-07-05T11:14:29.522287+00:00"},{"alias_kind":"pith_short_8","alias_value":"SDAS6GF2","created_at":"2026-07-05T11:14:29.522287+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":3,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.16517","citing_title":"How Post-Training Shapes Biological Reasoning Models","ref_index":72,"is_internal_anchor":false},{"citing_arxiv_id":"2512.15567","citing_title":"Evaluating Large Language Models in Scientific Discovery","ref_index":47,"is_internal_anchor":false},{"citing_arxiv_id":"2604.27351","citing_title":"Heterogeneous Scientific Foundation Model Collaboration","ref_index":40,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/SDAS6GF2L724EKGUIWJ2LKZ7HX","json":"https://pith.science/pith/SDAS6GF2L724EKGUIWJ2LKZ7HX.json","graph_json":"https://pith.science/api/pith-number/SDAS6GF2L724EKGUIWJ2LKZ7HX/graph.json","events_json":"https://pith.science/api/pith-number/SDAS6GF2L724EKGUIWJ2LKZ7HX/events.json","paper":"https://pith.science/paper/SDAS6GF2"},"agent_actions":{"view_html":"https://pith.science/pith/SDAS6GF2L724EKGUIWJ2LKZ7HX","download_json":"https://pith.science/pith/SDAS6GF2L724EKGUIWJ2LKZ7HX.json","view_paper":"https://pith.science/paper/SDAS6GF2","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2505.19501&json=true","fetch_graph":"https://pith.science/api/pith-number/SDAS6GF2L724EKGUIWJ2LKZ7HX/graph.json","fetch_events":"https://pith.science/api/pith-number/SDAS6GF2L724EKGUIWJ2LKZ7HX/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/SDAS6GF2L724EKGUIWJ2LKZ7HX/action/timestamp_anchor","attest_storage":"https://pith.science/pith/SDAS6GF2L724EKGUIWJ2LKZ7HX/action/storage_attestation","attest_author":"https://pith.science/pith/SDAS6GF2L724EKGUIWJ2LKZ7HX/action/author_attestation","sign_citation":"https://pith.science/pith/SDAS6GF2L724EKGUIWJ2LKZ7HX/action/citation_signature","submit_replication":"https://pith.science/pith/SDAS6GF2L724EKGUIWJ2LKZ7HX/action/replication_record"}},"created_at":"2026-07-05T11:14:29.522287+00:00","updated_at":"2026-07-05T11:14:29.522287+00:00"}