Recursive Environment Scaling for General Long-Horizon Agents

Lingxiao Jing, Zhengbo Jiao, Yichen Xiang, Bofan Zhu, Lingbei Meng, Yan Ni, Haorui Chen, Ran Ran, Zhirong Shen, Yuheng Yang, Chenhang Cui, Yaosong Du, Rui Huang

arXiv:2610.16397 · 2026-10-18 공개 · arXiv · PDF

benchmarking model-evaluation task-success long-horizon-agents dependency-composition recursive-environment-scaling trajectory-finetuning claw-eval-live

Abstract

We study environment scaling as a way to increase agent-benchmark difficulty while holding core task objectives fixed. Our framework, Recursive Environment Scaling (RES), transforms existing benchmark environments while preserving their objectives and completion requirements. A proposer changes information access, resource relations, and execution dependencies; execution-grounded checks validate the resulting candidate, and accepted changes are inherited across rounds. This produces related environments for paired evaluation and training-trajectory collection. Experiments on 294 initial tasks from three benchmarks show lower task success for all six evaluation models, with decreases of 57.14--77.14 percentage points on Claw-Eval-Live. With equal trajectory counts and identical optimization settings, fine-tuning on RES trajectories outperforms original-trajectory SFT in all nine model--benchmark comparisons by 7.00--14.61 percentage points. Ablations identify recursive inheritance and dependency composition as important contributors to increased difficulty. Together, these results support environment scaling for both agent evaluation and training.

한국어 요약

한 줄 요약

Recursive Environment Scaling(RES)은 기존 벤치마크 환경을 변형해 에이전트 평가와 학습을 강화하는 프레임워크로, Claw-Eval-Live에서 77.14% 성능 하락을 유도한다.

핵심 기여도

핵심 아이디어

기존 연구는 주로 태스크 생성이나 환경 합성에 집중했지만, RES는 기존 벤치마크 환경을 변형하면서 핵심 목표와 완료 조건을 유지하는 새로운 접근법을 제시한다. RES는 **proposer**가 자원과 실행 의존성을 변형하고, **execution-grounded checks**가 변형 후 환경의 유효성을 검증하며, 승인된 변화는 다음 라운드로 **재귀적으로 상속**된다. 이는 환경의 복잡도를 점진적으로 증가시키며, 에이전트가 동일한 목표를 달성하기 위해 더 많은 상호작용과 계획을 요구하게 만든다.

기술적 접근법

주요 결과

의의 및 한계

실용적 활용