DARS: Dynamic Action Re-Sampling to Enhance Coding Agent Performance by Adaptive Tree Traversal
This work addresses inefficiencies in coding agents for software engineering, offering a faster and more effective method to recover from errors, though it appears incremental as it builds on existing compute scaling strategies.
The paper tackles the problem of sub-optimal decision-making in LLM-powered coding agents by introducing Dynamic Action Re-Sampling (DARS), a novel inference time compute scaling approach that improves performance, achieving a pass@k score of 55% and a pass@1 rate of 47% on the SWE-Bench Lite benchmark.
Large Language Models (LLMs) have revolutionized various domains, including natural language processing, data analysis, and software development, by enabling automation. In software engineering, LLM-powered coding agents have garnered significant attention due to their potential to automate complex development tasks, assist in debugging, and enhance productivity. However, existing approaches often struggle with sub-optimal decision-making, requiring either extensive manual intervention or inefficient compute scaling strategies. To improve coding agent performance, we present Dynamic Action Re-Sampling (DARS), a novel inference time compute scaling approach for coding agents, that is faster and more effective at recovering from sub-optimal decisions compared to baselines. While traditional agents either follow linear trajectories or rely on random sampling for scaling compute, our approach DARS works by branching out a trajectory at certain key decision points by taking an alternative action given the history of the trajectory and execution feedback of the previous attempt from that point. We evaluate our approach on SWE-Bench Lite benchmark, demonstrating that this scaling strategy achieves a pass@k score of 55% with Claude 3.5 Sonnet V2. Our framework achieves a pass@1 rate of 47%, outperforming state-of-the-art (SOTA) open-source frameworks.