EnvOpt: Automated Optimization for Agent Environments
arXiv — cs.AI ·
What to know
The paper introduces EnvOpt, a method for automatically optimizing stationary environments (e.g., Slack workspaces, data collections) to reduce LLM agent costs and improve success rates, without needing graders or reference answers. On AutomationBench, EnvOpt reduced Claude Haiku 4.5's cost by 72.7% and raised its pass rate by 14.34%; the optimized environment transferred to Opus 5, cutting cost by 68.1% with a 1.34% gain. When combined with Meta-Harness's optimized harness for Haiku 4.5, EnvOpt further lowered cost by 83.4% and added 11.17% accuracy over Meta-Harness alone. Across 5 benchmarks, EnvOpt significantly reduced agent costs while improving or maintaining pass rates, and proved transferable across models and harnesses.
Source and evidence
arXiv — cs.AIOriginal language · und ·Original author: Mingxuan Li, Huanzhi Mao, Qiuyang Mang, Hanchen Li, Joseph E. Gonzalez, Alex Dimakis · Source: arXiv and the named paper authors. Descriptive metadata: CC0 1.0.Source licence
Why it matters by industry
- Artificial intelligence
This research introduces a novel optimization direction for AI agents—optimizing the environment rather than just model parameters or harnesses—with significant cost reductions and performance gains, relevant to AI model deployment and agent efficiency.
Reviewed summary. The complete article remains at its original source.