Search papers, labs, and topics across Lattice.
SWE-TRACE optimizes software engineering agents by first distilling a token-efficient SFT corpus using stepwise oracle verification, then training with a Rubric-Based Process Reward Model (PRM) that provides dense feedback on intermediate steps, and finally using the PRM for heuristic-guided Test-Time Scaling (TTS) to prune action candidates. This approach addresses challenges of unoptimized demonstration data, sparse rewards, and computationally expensive inference in long-horizon SWE tasks. Experiments on SWE benchmarks show SWE-TRACE improves resolution rates while reducing token consumption and inference latency.
Forget slow, bloated agents: SWE-TRACE slashes token consumption and latency while boosting software engineering task resolution by distilling efficient trajectories and using a rubric-based reward model for guided exploration and test-time scaling.
Resolving real-world software engineering (SWE) issues with autonomous agents requires complex, long-horizon reasoning. Current pipelines are bottlenecked by unoptimized demonstration data, sparse execution rewards, and computationally prohibitive inference scaling, which collectively exacerbate token bloat, reward hacking, and policy degradation. We present SWE-TRACE (Trajectory Reduction and Agentic Criteria Evaluation), a unified framework optimizing the SWE agent lifecycle across data curation, reinforcement learning (RL), and test-time inference. First, we introduce an LLM multi-task cascading method, utilizing stepwise oracle verification to distill a 60K-instance Supervised Fine-Tuning (SFT) corpus strictly biased toward token-efficient, shortest-path trajectories. Second, to overcome the instability of sparse outcome rewards, we design a MemoryAugmented Agentic RL pipeline featuring a Rubric-Based Process Reward Model (PRM). An auxiliary Rubric-Agent provides dense, fine-grained heuristic feedback on intermediate steps, guiding the model through long-horizon tasks. Finally, we bridge training and inference by repurposing the PRM for heuristic-guided Test-Time Scaling (TTS). By dynamically evaluating and pruning action candidates at each step, SWE-TRACE achieves superior search efficiency without the latency overhead of standard parallel sampling. Extensive experiments on standard SWE benchmarks demonstrate that SWE-TRACE significantly advances the state-of-the-art, maximizing resolution rates while drastically reducing both token consumption and inference latency.
