{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:E2JO7VS7MJEHH3DLMUOAFWYISV","short_pith_number":"pith:E2JO7VS7","schema_version":"1.0","canonical_sha256":"2692efd65f624873ec6b651c02db08955ce68135824a17e7f1e77a6e56a1b554","source":{"kind":"arxiv","id":"2602.02660","version":3},"attestation_state":"computed","paper":{"title":"MARS: Modular Agent with Reflective Search for Automated AI Research","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.AI","authors_text":"Bhavana Dalvi Mishra, Jaehyun Nam, Jiefeng Chen, Jinsung Yoon, Rui Meng, Tomas Pfister","submitted_at":"2026-02-02T19:00:03Z","abstract_excerpt":"A critical bottleneck in automating AI research is the execution of complex machine learning engineering (MLE) tasks. MLE differs from general software engineering due to computationally expensive evaluation (e.g., model training) and opaque performance attribution. Current LLM-based agents struggle here, often generating monolithic scripts that ignore execution costs and causal factors. We introduce MARS (Modular Agent with Reflective Search), a framework optimized for autonomous AI research. MARS relies on three pillars: (1) Budget-Aware Planning via cost-constrained Monte Carlo Tree Search "},"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":"2602.02660","kind":"arxiv","version":3},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.AI","submitted_at":"2026-02-02T19:00:03Z","cross_cats_sorted":[],"title_canon_sha256":"1650488212c618eacce2744d3b4ed5590cb2b2c42d99d2b2a7454e181f8ce814","abstract_canon_sha256":"8edb7a1eff1d42d6a8e04c3b8c168e8846a983fad75f9c8b2a68b2e60dc7663c"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-21T01:04:22.062361Z","signature_b64":"FAD1fR+iAJ5DE3aoY8AlqPY5WcQJKNJJV9L57CVASXDw3dpyniS/ZKO2VnHhh7c+OzL4nQbponBfedER/xubAQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"2692efd65f624873ec6b651c02db08955ce68135824a17e7f1e77a6e56a1b554","last_reissued_at":"2026-05-21T01:04:22.061809Z","signature_status":"signed_v1","first_computed_at":"2026-05-21T01:04:22.061809Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"MARS: Modular Agent with Reflective Search for Automated AI Research","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.AI","authors_text":"Bhavana Dalvi Mishra, Jaehyun Nam, Jiefeng Chen, Jinsung Yoon, Rui Meng, Tomas Pfister","submitted_at":"2026-02-02T19:00:03Z","abstract_excerpt":"A critical bottleneck in automating AI research is the execution of complex machine learning engineering (MLE) tasks. MLE differs from general software engineering due to computationally expensive evaluation (e.g., model training) and opaque performance attribution. Current LLM-based agents struggle here, often generating monolithic scripts that ignore execution costs and causal factors. We introduce MARS (Modular Agent with Reflective Search), a framework optimized for autonomous AI research. MARS relies on three pillars: (1) Budget-Aware Planning via cost-constrained Monte Carlo Tree Search "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2602.02660","kind":"arxiv","version":3},"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/2602.02660/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":"2602.02660","created_at":"2026-05-21T01:04:22.061888+00:00"},{"alias_kind":"arxiv_version","alias_value":"2602.02660v3","created_at":"2026-05-21T01:04:22.061888+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2602.02660","created_at":"2026-05-21T01:04:22.061888+00:00"},{"alias_kind":"pith_short_12","alias_value":"E2JO7VS7MJEH","created_at":"2026-05-21T01:04:22.061888+00:00"},{"alias_kind":"pith_short_16","alias_value":"E2JO7VS7MJEHH3DL","created_at":"2026-05-21T01:04:22.061888+00:00"},{"alias_kind":"pith_short_8","alias_value":"E2JO7VS7","created_at":"2026-05-21T01:04:22.061888+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":7,"internal_anchor_count":7,"sample":[{"citing_arxiv_id":"2605.17373","citing_title":"FML-bench: A Controlled Study of AI Research Agent Strategies from the Perspective of Search Dynamics","ref_index":10,"is_internal_anchor":true},{"citing_arxiv_id":"2605.15461","citing_title":"DrugSAGE:Self-evolving Agent Experience for Efficient State-of-the-Art Drug Discovery","ref_index":15,"is_internal_anchor":true},{"citing_arxiv_id":"2603.26499","citing_title":"AIRA_2: Overcoming Bottlenecks in AI Research Agents","ref_index":3,"is_internal_anchor":true},{"citing_arxiv_id":"2605.12913","citing_title":"Revisiting DAgger in the Era of LLM-Agents","ref_index":7,"is_internal_anchor":true},{"citing_arxiv_id":"2604.13018","citing_title":"Toward Autonomous Long-Horizon Engineering for ML Research","ref_index":3,"is_internal_anchor":true},{"citing_arxiv_id":"2604.12999","citing_title":"Agentic Discovery with Active Hypothesis Exploration for Visual Recognition","ref_index":8,"is_internal_anchor":true},{"citing_arxiv_id":"2604.14455","citing_title":"AIBuildAI: An AI Agent for Automatically Building AI Models","ref_index":25,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/E2JO7VS7MJEHH3DLMUOAFWYISV","json":"https://pith.science/pith/E2JO7VS7MJEHH3DLMUOAFWYISV.json","graph_json":"https://pith.science/api/pith-number/E2JO7VS7MJEHH3DLMUOAFWYISV/graph.json","events_json":"https://pith.science/api/pith-number/E2JO7VS7MJEHH3DLMUOAFWYISV/events.json","paper":"https://pith.science/paper/E2JO7VS7"},"agent_actions":{"view_html":"https://pith.science/pith/E2JO7VS7MJEHH3DLMUOAFWYISV","download_json":"https://pith.science/pith/E2JO7VS7MJEHH3DLMUOAFWYISV.json","view_paper":"https://pith.science/paper/E2JO7VS7","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2602.02660&json=true","fetch_graph":"https://pith.science/api/pith-number/E2JO7VS7MJEHH3DLMUOAFWYISV/graph.json","fetch_events":"https://pith.science/api/pith-number/E2JO7VS7MJEHH3DLMUOAFWYISV/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/E2JO7VS7MJEHH3DLMUOAFWYISV/action/timestamp_anchor","attest_storage":"https://pith.science/pith/E2JO7VS7MJEHH3DLMUOAFWYISV/action/storage_attestation","attest_author":"https://pith.science/pith/E2JO7VS7MJEHH3DLMUOAFWYISV/action/author_attestation","sign_citation":"https://pith.science/pith/E2JO7VS7MJEHH3DLMUOAFWYISV/action/citation_signature","submit_replication":"https://pith.science/pith/E2JO7VS7MJEHH3DLMUOAFWYISV/action/replication_record"}},"created_at":"2026-05-21T01:04:22.061888+00:00","updated_at":"2026-05-21T01:04:22.061888+00:00"}