Large Language Model (LLM)-based agents are increasingly employed to automate complex software engineering tasks, such as program repair and issue resolution. These agents operate by autonomously generating natural language thoughts, invoking external tools, and iteratively refining their solutions. Despite their widespread adoption, the internal decision-making processes of these agents remain largely unexplored, limiting our understanding of their operational dynamics and failure modes. In this paper, we present a large-scale empirical study of the thought-action-result trajectories of three state-of-the-art LLM-based agents: RepairAgent, AutoCodeRover, and OpenHands. We unify their interaction logs into a common format, capturing 120 trajectories and 2,822 LLM interactions focused on program repair and issue resolution. Our study combines quantitative analyses of structural properties, action patterns, and token usage with qualitative assessments of reasoning coherence and feedback integration. We identify key trajectory characteristics, such as iteration counts and token consumption, recurring action sequences, and the semantic coherence of thoughts, actions, and their results. Our findings reveal behavioral motifs and anti-patterns that distinguish successful from failed executions, providing actionable insights for improving agent design, including prompting strategies, failure diagnosis, and anti-pattern detection. We release our dataset and annotation framework to support further research on transparent and robust autonomous software engineering agents.

Large Language Models (LLMs) have recently shown strong potential in automatic program repair (APR), especially in repository-level settings where the goal is to generate patches based on natural language issue descriptions, large codebases, and regression tests. However, despite their promise, current LLM-based APR techniques often struggle to produce correct fixes due to limited understanding of code context and over-reliance on incomplete test suites. As a result, they frequently generate Draft Patches-partially correct patches that either incompletely address the bug or overfit to the test cases. In this work, we propose a novel patch refinement framework, Refine, that systematically transforms Draft Patches into correct ones. Refine addresses three key challenges: disambiguating vague issue and code context, diversifying patch candidates through test-time scaling, and aggregating partial fixes via an LLM-powered code review process. We implement Refine as a general refinement module that can be integrated into both open-agent-based and workflow-based APR systems. Our evaluation on the SWE-Bench Lite benchmark shows that Refine achieves state-of-the-art results among workflow-based approaches and approaches the best-known performance across all APR categories. Specifically, Refine boosts AutoCodeRover's performance by 14.67%, achieving a score of 51.67% and surpassing all prior baselines. On SWE-Bench Verified, Refine improves the resolution rate by 12.2%, and when integrated across multiple APR systems, it yields an average improvement of 14%-demonstrating its broad effectiveness and generalizability. These results highlight the effectiveness of refinement as a missing component in current APR pipelines and the potential of agentic collaboration in closing the gap between near-correct and correct patches. We also open source our code.

AI agents have recently shown significant promise in software engineering. Much public attention has been transfixed on the topic of code generation from Large Language Models (LLMs) via a prompt. However, software engineering is much more than programming, and AI agents go far beyond instructions given by a prompt. At the code level, common software tasks include code generation, testing, and program repair. Design level software tasks may include architecture exploration, requirements understanding, and requirements enforcement at the code level. Each of these software tasks involves micro-decisions which can be taken autonomously by an AI agent, aided by program analysis tools. This creates the vision of an AI software engineer, where the AI agent can be seen as a member of a development team. Conceptually, the key to successfully developing trustworthy agentic AI-based software workflows will be to resolve the core difficulty in software engineering - the deciphering and clarification of developer intent. Specification inference, or deciphering the intent, thus lies at the heart of many software tasks, including software maintenance and program repair. A successful deployment of agentic technology into software engineering would involve making conceptual progress in such intent inference via agents. Trusting the AI agent becomes a key aspect, as software engineering becomes more automated. Higher automation also leads to higher volume of code being automatically generated, and then integrated into code-bases. Thus to deal with this explosion, an emerging direction is AI-based verification and validation (V & V) of AI generated code. We posit that agentic software workflows in future will include such AIbased V&V.

The advancement of large language models (LLMs) and code agents has demonstrated significant potential to assist software engineering (SWE) tasks, such as autonomous issue resolution and feature addition. Existing AI for software engineering leaderboards (e.g., SWE-bench) focus solely on solution accuracy, ignoring the crucial factor of effectiveness in a resource-constrained world. This is a universal problem that also exists beyond software engineering tasks: any AI system should be more than correct - it must also be cost-effective. To address this gap, we introduce SWE-Effi, a set of new metrics to re-evaluate AI systems in terms of holistic effectiveness scores. We define effectiveness as the balance between the accuracy of outcome (e.g., issue resolve rate) and the resources consumed (e.g., token and time). In this paper, we specifically focus on the software engineering scenario by re-ranking popular AI systems for issue resolution on a subset of the SWE-bench benchmark using our new multi-dimensional metrics. We found that AI system's effectiveness depends not just on the scaffold itself, but on how well it integrates with the base model, which is key to achieving strong performance in a resource-efficient manner. We also identified systematic challenges such as the "token snowball" effect and, more significantly, a pattern of "expensive failures". In these cases, agents consume excessive resources while stuck on unsolvable tasks - an issue that not only limits practical deployment but also drives up the cost of failed rollouts during RL training. Lastly, we observed a clear trade-off between effectiveness under the token budget and effectiveness under the time budget, which plays a crucial role in managing project budgets and enabling scalable reinforcement learning, where fast responses are essential.

We introduce CTIM-Rover, an AI agent for Software Engineering (SE) built on top of AutoCodeRover (Zhang et al., 2024) that extends agentic reasoning frameworks with an episodic memory, more specifically, a general and repository-level Cross-Task-Instance Memory (CTIM). While existing open-source SE agents mostly rely on ReAct (Yao et al., 2023b), Reflexion (Shinn et al., 2023), or Code-Act (Wang et al., 2024), all of these reasoning and planning frameworks inefficiently discard their long-term memory after a single task instance. As repository-level understanding is pivotal for identifying all locations requiring a patch for fixing a bug, we hypothesize that SE is particularly well positioned to benefit from CTIM. For this, we build on the Experiential Learning (EL) approach ExpeL (Zhao et al., 2024), proposing a Mixture-Of-Experts (MoEs) inspired approach to create both a general-purpose and repository-level CTIM. We find that CTIM-Rover does not outperform AutoCodeRover in any configuration and thus conclude that neither ExpeL nor DoT-Bank (Lingam et al., 2024) scale to real-world SE problems. Our analysis indicates noise introduced by distracting CTIM items or exemplar trajectories as the likely source of the performance degradation.

As foundation models (FMs) play an increasingly prominent role in complex software systems, such as FM-powered agentic software (i.e., Agentware), they introduce significant challenges for developers regarding observability. Unlike traditional software, agents operate autonomously, using extensive data and opaque implicit reasoning, making it difficult to observe and understand their behavior during runtime, especially when they take unexpected actions or encounter errors. In this paper, we highlight the limitations of traditional operational observability in the context of FM-powered software, and introduce cognitive observability as a new type of required observability that has emerged for such innovative systems. We then propose a novel framework that provides cognitive observability into the implicit reasoning processes of agents (a.k.a. reasoning observability), and demonstrate the effectiveness of our framework in boosting the debuggability of Agentware and, in turn, the abilities of an Agentware through a case study on AutoCodeRover, a cuttingedge Agentware for autonomous program improvement.

Recent advancements in code-fluent Large Language Models (LLMs) enabled the research on repository-level code editing. In such tasks, the model navigates and modifies the entire codebase of a project according to request. Hence, such tasks require efficient context retrieval, i.e., navigating vast codebases to gather relevant context. Despite the recognized importance of context retrieval, existing studies tend to approach repository-level coding tasks in an end-to-end manner, rendering the impact of individual components within these complicated systems unclear. In this work, we decouple the task of context retrieval from the other components of the repository-level code editing pipelines. We lay the groundwork to define the strengths and weaknesses of this component and the role that reasoning plays in it by conducting experiments that focus solely on context retrieval. We conclude that while the reasoning helps to improve the precision of the gathered context, it still lacks the ability to identify its sufficiency. We also outline the ultimate role of the specialized tools in the process of context gathering. The code supplementing this paper is available at https://github.com/JetBrains-Research/ai-agents-code-editing.

Researchers have made significant progress in automating the software development process in the past decades. Recent progress in Large Language Models (LLMs) has significantly impacted the development process, where developers can use LLM-based programming assistants to achieve automated coding. Nevertheless software engineering involves the process of program improvement apart from coding, specifically to enable software maintenance (e.g. bug fixing) and software evolution (e.g. feature additions). In this paper, we propose an automated approach for solving GitHub issues to autonomously achieve program improvement. In our approach called AutoCodeRover, LLMs are combined with sophisticated code search capabilities, ultimately leading to a program modification or patch. In contrast to recent LLM agent approaches from AI researchers and practitioners, our outlook is more software engineering oriented. We work on a program representation (abstract syntax tree) as opposed to viewing a software project as a mere collection of files. Our code search exploits the program structure in the form of classes/methods to enhance LLM's understanding of the issue's root cause, and effectively retrieve a context via iterative search. The use of spectrum based fault localization using tests, further sharpens the context, as long as a test-suite is available. Experiments on SWE-bench-lite which consists of 300 real-life GitHub issues show increased efficacy in solving GitHub issues (22-23% on SWE-bench-lite). On the full SWE-bench consisting of 2294 GitHub issues, AutoCodeRover solved around 16% of issues, which is higher than the efficacy of the recently reported AI software engineer Devin from Cognition Labs, while taking time comparable to Devin. We posit that our workflow enables autonomous software engineering, where, in future, auto-generated code from LLMs can be autonomously improved.