{"paper":{"title":"Toward Autonomous Long-Horizon Engineering for ML Research","license":"http://creativecommons.org/licenses/by/4.0/","headline":"AiScientist achieves higher performance on long-horizon ML research benchmarks by using hierarchical orchestration and a File-as-Bus workspace.","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Cheng Chen, Fanzhe Meng, Guoxin Chen, Jiale Zhao, Jie Chen, Ji-Rong Wen, Kai Jia, Lei Chen, Ruihua Song, Wayne Xin Zhao","submitted_at":"2026-04-14T17:55:16Z","abstract_excerpt":"Agentic systems increasingly automate pieces of AI research. Yet turning underspecified research objectives into runnable, experimentally validated ML systems remains a central bottleneck. We study this operational setting as \\emph{long-horizon ML research engineering}: converting a research specification into a runnable ML system through repeated implementation, experimentation, and refinement. The central challenge is to sustain cumulative project progress across heterogeneous stages under delayed, confounded feedback. We introduce AiScientist, a multi-agent system built around thin control "},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"AiScientist improves PaperBench score by 10.54 points on average over the best matched baseline and achieves 81.82 Any Medal% on MLE-Bench Lite. Ablation studies further show that File-as-Bus protocol is a key driver of performance, reducing PaperBench by 6.41 points and MLE-Bench Lite by 31.82 points when removed.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"That the chosen benchmarks (PaperBench and MLE-Bench Lite) accurately capture real long-horizon ML research engineering capability and that the reported gains are attributable to the hierarchical orchestration plus File-as-Bus design rather than implementation details or baseline mismatches.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"AiScientist improves ML research benchmarks by 10.54 points on PaperBench and reaches 81.82% Any Medal on MLE-Bench Lite through hierarchical control plus durable file-based state instead of conversational handoffs.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"AiScientist achieves higher performance on long-horizon ML research benchmarks by using hierarchical orchestration and a File-as-Bus workspace.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"55774e1c0ae4bdfc0eb6b15246c9d281f0c6014f2be0d8c89ebe2267b02461ac"},"source":{"id":"2604.13018","kind":"arxiv","version":2},"verdict":{"id":"937606f7-c438-4d2e-b510-61a6625585f5","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-10T15:46:53.395337Z","strongest_claim":"AiScientist improves PaperBench score by 10.54 points on average over the best matched baseline and achieves 81.82 Any Medal% on MLE-Bench Lite. Ablation studies further show that File-as-Bus protocol is a key driver of performance, reducing PaperBench by 6.41 points and MLE-Bench Lite by 31.82 points when removed.","one_line_summary":"AiScientist improves ML research benchmarks by 10.54 points on PaperBench and reaches 81.82% Any Medal on MLE-Bench Lite through hierarchical control plus durable file-based state instead of conversational handoffs.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"That the chosen benchmarks (PaperBench and MLE-Bench Lite) accurately capture real long-horizon ML research engineering capability and that the reported gains are attributable to the hierarchical orchestration plus File-as-Bus design rather than implementation details or baseline mismatches.","pith_extraction_headline":"AiScientist achieves higher performance on long-horizon ML research benchmarks by using hierarchical orchestration and a File-as-Bus workspace."},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2604.13018/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"}