AI summaryⓘ
The authors created AiScientist, a system that helps AI work on complex machine learning research projects over long periods, like hours or days. Instead of just chatting back and forth between parts, their system uses a special workspace where important files like plans and results are saved and shared, so different parts remember everything clearly. This approach helps AiScientist organize tasks better and keep track of progress, leading to better results on testing benchmarks. Their experiments show that keeping this shared file system is really important for success. Overall, the authors suggest that long-term AI research needs good teamwork around solid shared information, not just quick thinking in small steps.
Autonomous AILong-horizon ML researchHierarchical orchestrationFile-as-Bus workspaceDurable state continuityMachine learning experimentationBenchmarkingProject state management
Authors
Guoxin Chen, Jie Chen, Lei Chen, Jiale Zhao, Fanzhe Meng, Wayne Xin Zhao, Ruihua Song, Cheng Chen, Ji-Rong Wen, Kai Jia
Abstract
Autonomous AI research has advanced rapidly, but long-horizon ML research engineering remains difficult: agents must sustain coherent progress across task comprehension, environment setup, implementation, experimentation, and debugging over hours or days. We introduce AiScientist, a system for autonomous long-horizon engineering for ML research built on a simple principle: strong long-horizon performance requires both structured orchestration and durable state continuity. To this end, AiScientist combines hierarchical orchestration with a permission-scoped File-as-Bus workspace: a top-level Orchestrator maintains stage-level control through concise summaries and a workspace map, while specialized agents repeatedly re-ground on durable artifacts such as analyses, plans, code, and experimental evidence rather than relying primarily on conversational handoffs, yielding thin control over thick state. Across two complementary benchmarks, 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. These results suggest that long-horizon ML research engineering is a systems problem of coordinating specialized work over durable project state, rather than a purely local reasoning problem.