Dong Sun

Jiawei He, Jie Jia, Chenbo Liu et al.

Existing benchmarks for LLM coding agents mainly evaluate final outcomes, such as task completion, compilation success, and test pass rates. While these metrics are useful for measuring end-task capability, they provide limited visibility into how an execution unfolds and often miss recurrent process-level failures that arise during multi-step operation. We present ProcBench, a benchmark-oriented framework for evaluating coding-agent trajectories through process defects and control preservation. ProcBench organizes execution failures into a reusable ontology, standardizes heterogeneous logs into a unified trajectory representation, and reports calibrated risk-based scorecards instead of relying only on final outcomes. We instantiate ProcBench on an annotated set of 200 trajectories and apply it across three coding-agent benchmarks: AndroidBench, TerminalBench, and SWE-bench-Verified. Our results suggest that ProcBench can be instantiated with useful reliability, that calibration improves the empirical interpretability of defect findings relative to direct thresholding, and that process-aware scorecards provide diagnostic distinctions beyond conventional outcome-based evaluation. We also discuss limitations, including annotation dependence, partial observability for some defect classes, and the need for broader external validation.

Jiawei He, Weisong Sun, Mengyu Shi et al.

Large language models have shown strong performance on software engineering (SE) tasks, yet understanding large industrial repositories remains challenging. Existing methods often retrieve only local fragments and fail to recover the broader task-relevant context needed for complex repository-level tasks. We present DeepDiscovery, a task-level repository-understanding method for large industrial codebases. DeepDiscovery uses a two-stage \textit{Location--Inference} framework to localize high-confidence task anchors and recover broader task-relevant context over multi-relational repository structure under budget constraints. Across controlled method-level evaluation, organization-internal industrial repository-understanding scenarios, and end-to-end evaluation on SWE-bench Verified, DeepDiscovery consistently improves task-relevant file recovery and downstream SE performance. On 27 medium-scale tasks, DeepDiscovery achieves the best file recovery quality among five representative baselines without offline preprocessing. On organization-internal industrial tasks from a production-scale integrated codebase ecosystem, including 27 medium-scale tasks and 40 large-scale tasks, DeepDiscovery improves Full Recall Rate across multiple AI coding systems, with absolute gains ranging from 1.6 to 9.2 percentage points on large subprojects and from 2.5 to 7.4 percentage points on medium-scale subprojects. In a controlled end-to-end evaluation on SWE-bench Verified, a system equipped with DeepDiscovery achieves a 78.6\% Solve Rate, outperforming the corresponding baseline by 8.2 percentage points. These results suggest that stronger task-level repository understanding can improve coding-agent performance on complex SE tasks.