Stateless language agents improve long running automated research efficiency
Stateless Language Agents: Scaling Long-Horizon Automated Research
Machine LearningArtificial IntelligenceComputation and Language
Summary
When using language models to automate research, simply running them longer doesn't always lead to better results because they can repeat themselves or waste effort. The authors show that this happens because the agents keep growing conversation histories and decide what to try next on their own. They introduce Stateless Language Agents (SLAs), which keep all progress outside the agents and give them fresh information each time, focusing their work better. This new method uses less than 16% of the tokens needed by prior systems to reach similar performance and achieves better results overall.
What this means in practice
- •For software engineering teams: Improve software debugging and optimization by running stateless language agents that assign focused experiments efficiently with less compute.
- •For hardware performance engineers: Use stateless language agents to speed up kernel optimization by keeping state separate and directing focused search for better results with fewer tokens.
Authors
Qizheng Zhang, Changxiu Ji, Isaac Sun, Yuetai Li, Shubhangi Upasani, Sherry Ruan, Boyuan Ma, Fenglu Hong, Vamsidhar Kamanuru, Yoonho Lee, Yuzhen Mao, Genghan Zhang, Rulin Shao, Qiuyang Mang, Andy Dimnaku, Changran Hu, Radha Poovendran, Kunle Olukotun
Abstract
Automated research systems increasingly run LLM agents over long horizons, but more inference does not by itself produce more progress: agents replay growing histories, duplicate one another's work, or stop experimenting while token consumption continues. Yet most evaluations use short budgets or benchmarks that saturate early, leaving these failure modes untested. We trace these failures to two choices: where research state lives and who decides what to try next. We introduce Stateless Language Agents (SLAs), built on the principle of stateful search with stateless agents: no agent carries its conversation across invocations; instead, the harness owns the research state (candidate solutions and measured outcomes) and reconstructs a fresh and role-specific context for every invocation. What each agent sees becomes an explicit design choice rather than a history that grows with the run. We implement this principle in the SLA framework, where a stateless Advisor reads harness-summarized evidence across search directions and assigns concrete experiments to parallel Workers. We evaluate SLA against three recent frameworks on software engineering, kernel optimization, and algorithm design at budgets of up to one billion tokens. SLA achieves the best final result on every task and reaches the strongest kernel baseline's final performance with over 84% fewer tokens. Ablations from shared checkpoints show that focused contexts and explicit assignments each contribute to SLA's progress, with effects that can compound over full runs, while the Advisor consumes less than 0.6% of tokens. These results argue for SLAs, which keep durable research state out of agent conversations, and show that short evaluation horizons can misjudge research systems and their components.