CTA framework detects 522 skill influence patterns in LLM agent traces across 49 tasks where average pass rate shifts only +0.3%, exposing evaluation gaps in behavioral effects like template copying and excess planning.
Beyond Final Code: A Process-Oriented Error Analysis of Software Development Agents in Real-World GitHub Scenarios
5 Pith papers cite this work. Polarity classification is still indexing.
5
Pith papers citing it
abstract
AI-driven software development has rapidly advanced with the emergence of software development agents that leverage large language models (LLMs) to tackle complex, repository-level software engineering tasks. These agents go beyond just generation of final code; they engage in multi-step reasoning, utilize various tools for code modification and debugging, and interact with execution environments to diagnose and iteratively resolve issues. However, most existing evaluations focus primarily on static analyses of final code outputs, yielding limited insights into the agents' dynamic problem-solving processes. To fill this gap, we conduct an in-depth empirical study on 3,977 solving-phase trajectories and 3,931 testing-phase logs from 8 top-ranked agents evaluated on 500 GitHub issues in the SWE-Bench benchmark. Our exploratory analysis shows that Python execution errors during the issue resolution phase correlate with lower resolution rates and increased reasoning overheads. We have identified the most prevalent errors -- such as ModuleNotFoundError and TypeError -- and highlighted particularly challenging errors like OSError and database-related issues (e.g., IntegrityError) that demand significantly more debugging effort. Furthermore, we have discovered 3 bugs in the SWE-Bench platform that affect benchmark fairness and accuracy; these issues have been reported to and confirmed by the maintainers. To promote transparency and foster future research, we publicly share our datasets and analysis scripts.
years
2026 5representative citing papers
PROBE structures runtime telemetry into diagnoses and evidence-grounded guidance, raising recovery rates by 12.45 points over baselines on 257 unresolved software repair and AIOps cases.
Large-scale trajectory analysis of 19 coding agents on 500 tasks finds that LLM choice drives outcomes more than framework design and that context-gathering plus validation behaviors improve success beyond task difficulty predictions.
Introduces the first benchmark for Java reproduction test generation from repository issues and adapts a prior Python tool to produce high performance on it.
An external controller for frozen LLMs raises strict validation success on three RL coding tasks from 0/9 to 8/9 by selecting memory records and skills, running fail-fast checks, and propagating credit via eligibility traces.
citing papers explorer
-
Counterfactual Trace Auditing of LLM Agent Skills
CTA framework detects 522 skill influence patterns in LLM agent traces across 49 tasks where average pass rate shifts only +0.3%, exposing evaluation gaps in behavioral effects like template copying and excess planning.
-
Debugging the Debuggers: Failure-Anchored Structured Recovery for Software Engineering Agents
PROBE structures runtime telemetry into diagnoses and evidence-grounded guidance, raising recovery rates by 12.45 points over baselines on 257 unresolved software repair and AIOps cases.
-
Beyond Resolution Rates: Behavioral Drivers of Coding Agent Success and Failure
Large-scale trajectory analysis of 19 coding agents on 500 tasks finds that LLM choice drives outcomes more than framework design and that context-gathering plus validation behaviors improve success beyond task difficulty predictions.
-
Reproduction Test Generation for Java SWE Issues
Introduces the first benchmark for Java reproduction test generation from repository issues and adapts a prior Python tool to produce high performance on it.
-
PYTHALAB-MERA: Validation-Grounded Memory, Retrieval, and Acceptance Control for Frozen-LLM Coding Agents
An external controller for frozen LLMs raises strict validation success on three RL coding tasks from 0/9 to 8/9 by selecting memory records and skills, running fail-fast checks, and propagating credit via eligibility traces.