{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:43XXVBN7B2STWAFEOLIGEYPYGF","short_pith_number":"pith:43XXVBN7","schema_version":"1.0","canonical_sha256":"e6ef7a85bf0ea53b00a472d06261f83140a4ffbe798a54245d8619dc6b86b7cd","source":{"kind":"arxiv","id":"2506.02314","version":1},"attestation_state":"computed","paper":{"title":"ResearchCodeBench: Benchmarking LLMs on Implementing Novel Machine Learning Research Code","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.CL"],"primary_cat":"cs.AI","authors_text":"Benjamin Klieger, Fan-Yun Sun, Harper Hua, Nick Haber, Sang T. Truong, Tianyu Hua, Violet Xiang, Weixin Liang","submitted_at":"2025-06-02T23:04:12Z","abstract_excerpt":"Large language models (LLMs) have shown promise in transforming machine learning research, yet their capability to faithfully implement novel ideas from recent research papers-ideas unseen during pretraining-remains unclear. We introduce ResearchCodeBench, a benchmark of 212 coding challenges that evaluates LLMs' ability to translate cutting-edge ML contributions from top 2024-2025 research papers into executable code. We assessed 30+ proprietary and open-source LLMs, finding that even the best models correctly implement less than 40% of the code. We find Gemini-2.5-Pro-Preview to perform best"},"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":"2506.02314","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.AI","submitted_at":"2025-06-02T23:04:12Z","cross_cats_sorted":["cs.CL"],"title_canon_sha256":"a51ab7df95b1fceceb1387dfb597ae0d8e8b7b63b60fd080662603a6669a1ac8","abstract_canon_sha256":"3f1d99f0507ea5a51dc23ff600484b362e797c1dde69f2ba6869a3d5e5cd424f"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:14:36.253173Z","signature_b64":"F5jF7r7U4gBXc7wfPch2MMlvA5IKJwGNYNWjSejdMmgV5hdrptOgeoDjJkLBggFUQVUDlRu8up9IrUT9A+FVBA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"e6ef7a85bf0ea53b00a472d06261f83140a4ffbe798a54245d8619dc6b86b7cd","last_reissued_at":"2026-07-05T11:14:36.252691Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:14:36.252691Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"ResearchCodeBench: Benchmarking LLMs on Implementing Novel Machine Learning Research Code","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.CL"],"primary_cat":"cs.AI","authors_text":"Benjamin Klieger, Fan-Yun Sun, Harper Hua, Nick Haber, Sang T. Truong, Tianyu Hua, Violet Xiang, Weixin Liang","submitted_at":"2025-06-02T23:04:12Z","abstract_excerpt":"Large language models (LLMs) have shown promise in transforming machine learning research, yet their capability to faithfully implement novel ideas from recent research papers-ideas unseen during pretraining-remains unclear. We introduce ResearchCodeBench, a benchmark of 212 coding challenges that evaluates LLMs' ability to translate cutting-edge ML contributions from top 2024-2025 research papers into executable code. We assessed 30+ proprietary and open-source LLMs, finding that even the best models correctly implement less than 40% of the code. We find Gemini-2.5-Pro-Preview to perform best"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2506.02314","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/2506.02314/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":"2506.02314","created_at":"2026-07-05T11:14:36.252751+00:00"},{"alias_kind":"arxiv_version","alias_value":"2506.02314v1","created_at":"2026-07-05T11:14:36.252751+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2506.02314","created_at":"2026-07-05T11:14:36.252751+00:00"},{"alias_kind":"pith_short_12","alias_value":"43XXVBN7B2ST","created_at":"2026-07-05T11:14:36.252751+00:00"},{"alias_kind":"pith_short_16","alias_value":"43XXVBN7B2STWAFE","created_at":"2026-07-05T11:14:36.252751+00:00"},{"alias_kind":"pith_short_8","alias_value":"43XXVBN7","created_at":"2026-07-05T11:14:36.252751+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":10,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2605.28371","citing_title":"From paper to benchmark: agentic, framework-based reproduction of under-specified methods in machine health intelligence","ref_index":12,"is_internal_anchor":false},{"citing_arxiv_id":"2512.01089","citing_title":"CodeDistiller: Automatically Generating Code Libraries for Scientific Coding Agents","ref_index":7,"is_internal_anchor":false},{"citing_arxiv_id":"2605.16616","citing_title":"MLReplicate: Benchmarking Autonomous Research Systems for Machine Learning Reproducibility","ref_index":13,"is_internal_anchor":false},{"citing_arxiv_id":"2605.18661","citing_title":"AI for Auto-Research: Roadmap & User Guide","ref_index":71,"is_internal_anchor":false},{"citing_arxiv_id":"2605.17324","citing_title":"ASPI: Seeking Ambiguity Clarification Amplifies Prompt Injection Vulnerability in LLM Agents","ref_index":73,"is_internal_anchor":false},{"citing_arxiv_id":"2605.15226","citing_title":"Is Agentic AI Ready for Real-World Hardware Engineering? A Deep Dive with Phoenix-bench","ref_index":33,"is_internal_anchor":false},{"citing_arxiv_id":"2602.11354","citing_title":"ReplicatorBench: Benchmarking LLM Agents for Replicability in Social and Behavioral Sciences","ref_index":7,"is_internal_anchor":false},{"citing_arxiv_id":"2604.04074","citing_title":"FactReview: Evidence-Grounded Reviews with Literature Positioning and Execution-Based Claim Verification","ref_index":4,"is_internal_anchor":false},{"citing_arxiv_id":"2604.19606","citing_title":"AblateCell: A Reproduce-then-Ablate Agent for Virtual Cell Repositories","ref_index":42,"is_internal_anchor":false},{"citing_arxiv_id":"2604.10718","citing_title":"SciPredict: Can LLMs Predict the Outcomes of Scientific Experiments in Natural Sciences?","ref_index":18,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/43XXVBN7B2STWAFEOLIGEYPYGF","json":"https://pith.science/pith/43XXVBN7B2STWAFEOLIGEYPYGF.json","graph_json":"https://pith.science/api/pith-number/43XXVBN7B2STWAFEOLIGEYPYGF/graph.json","events_json":"https://pith.science/api/pith-number/43XXVBN7B2STWAFEOLIGEYPYGF/events.json","paper":"https://pith.science/paper/43XXVBN7"},"agent_actions":{"view_html":"https://pith.science/pith/43XXVBN7B2STWAFEOLIGEYPYGF","download_json":"https://pith.science/pith/43XXVBN7B2STWAFEOLIGEYPYGF.json","view_paper":"https://pith.science/paper/43XXVBN7","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2506.02314&json=true","fetch_graph":"https://pith.science/api/pith-number/43XXVBN7B2STWAFEOLIGEYPYGF/graph.json","fetch_events":"https://pith.science/api/pith-number/43XXVBN7B2STWAFEOLIGEYPYGF/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/43XXVBN7B2STWAFEOLIGEYPYGF/action/timestamp_anchor","attest_storage":"https://pith.science/pith/43XXVBN7B2STWAFEOLIGEYPYGF/action/storage_attestation","attest_author":"https://pith.science/pith/43XXVBN7B2STWAFEOLIGEYPYGF/action/author_attestation","sign_citation":"https://pith.science/pith/43XXVBN7B2STWAFEOLIGEYPYGF/action/citation_signature","submit_replication":"https://pith.science/pith/43XXVBN7B2STWAFEOLIGEYPYGF/action/replication_record"}},"created_at":"2026-07-05T11:14:36.252751+00:00","updated_at":"2026-07-05T11:14:36.252751+00:00"}