The authors designed TQS (Trajectory Quality Score), a framework that scores agent trajectories based on structured convergence, operationalized through seven linear-time measures covering information focus, intent-action alignment, and policy dynamics.
Trajectory process quality, not just outcome, improves coding-agent fine-tuning
The authors propose TQS, a linear-time structural scorer that ranks successful agent trajectories and shows that fine-tuning on top-ranked ones beats random or bottom-ranked subsets on SWE-Bench-Verified.
Chinese Tech
Sida Li · Yuanxin Liu · Shicheng Li · Linli Yao · Xu Sun · Haoli Bai · +1 more
Peking University · Huawei Technologies Co., Ltd.
Research Digest··2 min read
The authors introduce TQS, a structural scoring framework that evaluates coding-agent trajectory quality from event logs alone, without LLM judges or reward models.
Why this paper
From Huawei Technologies Co., Ltd. and Peking University
In one line
Structural trajectory-quality scoring, not patch success, determines which coding-agent trajectories improve supervised fine-tuning.
What we could check
- ·No code link found
- ·No weights link found
- ·No dataset link found
- ·No compute details found
- ·No stated limitations found
- ·No benchmark numbers found
Observed from the paper text and links we have. Absence here means we did not find it, not that it does not exist.
§