LS-CM (Long-Short Collaborative Memory), a dual-memory architecture that couples a Retrieval-Augmented Generation (RAG)-based short-term memory for code-base context with a Reinforcement Learning (RL)driven long-term memory that distills strategic patterns from historical test failures and patch iterations, is proposed.
Abstract
Large Language Model (LLM)-based autonomous agents have shown significant promise in automating software engineering tasks, yet existing systems still suffer from two fundamental limitations: i) the lack of persistent experiential knowledge across debugging sessions, which forces agents to repeat exploratory mistakes, and ii) the static nature of prompt structures, which prevents agents from adapting their reasoning strategy to recurring failure modes. To address these gaps, we propose LS-CM (Long-Short Collaborative Memory), a dual-memory architecture that couples a Retrieval-Augmented Generation (RAG)-based short-term memory for code-base context with a Reinforcement Learning (RL)-driven long-term memory that distills strategic patterns from historical test failures and patch iterations. The core innovation is a self-evolving context engineering module that automatically refines the agent’s internal prompt structure based on execution feedback signals, treating the prompt as a configurable policy object that a gradient-free controller selects among pre-generated variants, rather than a fixed artifact. We evaluate LS-CM on two industry-standard benchmarks: SWE-bench, for end-to-end issue resolution on real-world GitHub repositories, and Defects4J, for bug localization and program repair in Java projects. Across multiple LLM backbones, LS-CM resolves 33.7% of SWE-bench Lite issues, a modest gain over the strongest reproduced baseline (Agentless, 32.0%), and improves Top-1 fault localization accuracy on Defects4J from 42.6% to 47.6% while reducing editing churn by 31%. We find that LS-CM’s most robust advantages lie in debugging efficiency, cross-episode strategy transfer, and run-to-run stability rather than in a single headline resolved-rate number. Ablation studies confirm that short-term retrieval, long-term policy learning, and self-evolving prompts each contribute to the gains. Our work demonstrates that explicitly modeling memory and prompt adaptation as first-class learning objects substantially enhances agent productivity on real-world software engineering tasks.
PMCoder is presented, an issue-resolution agent that couples a hierarchical phase planner with episodic memory that outperforms either component alone and reduces repeated failed actions, empty-patch exits, and context-window exhaustion.
TDD-Agent is introduced, which operationalizes the test-driven development paradigm for code generation and improves not only code correctness but also the effectiveness of the generated tests, yielding higher pass rates, coverage, and mutation scores, suggesting that tests can serve as evolving reasoning artifacts rather than fixed validators.
Hong Yu, Ke-Fan Li, Jia-Kun Li et al.· 2 citations
This report synthesizes peer-reviewed and archival research published between 2019 and 2024 on the use of large language models and multi-agent architectures for automated vulnerability repair, with a specific focus on how these techniques apply, or fail to apply, to legacy software systems. The review traces a clear methodological progression across this five-year window: from template and heuristic program repair, through neural and transfer-learning models, to fine-tuned transformer-based repair, zero-shot and conversational prompting of general-purpose LLMs, and finally to multi-agent pipelines in which specialized agents divide the work of detecting, localizing, patching, and validating a vulnerability. Two figures accompany the discussion. The first depicts a generalized multi-agent repair pipeline synthesized from the architectures described across the reviewed literature, showing how detection, localization, patch generation, and validation agents typically interact, including the validation feedback loop that distinguishes conversational and agentic systems from earlier single-shot approaches. The second traces the field's methodological evolution chronologically, from 2019's template-based baselines through 2024's agentic systems. A comparative table consolidates fifteen of the most relevant studies discussed in the body of the report, listing the approach, target language, and publication venue for each. The report concludes that while multi-agent LLM pipelines show real promise for legacy code specifically, since their capacity for autonomous context reconstruction and iterative, feedback-driven refinement directly addresses the sparse documentation and thin test coverage that define legacy environments, the literature to date has been validated almost exclusively on actively maintained, comparatively well-organized open-source repositories. The gap between these benchmarks and the scale and disorganization of real enterprise legacy systems has not yet been closed empirically, and closing it, together with building trust and explainability mechanisms suitable for security-critical maintenance work, is identified as the field's most consequential open problem.
Harsh Verma· International Journal of Sci...· 0 citations
This survey aims to clarify the conceptual boundaries of self-evolving coding agents and provide a foundation for designing more adaptive, reliable, and software-aware agentic systems.
H. Zhou, Haichuan Hu, Tianyu Luo et al.· 0 citations
Large language model (LLM)-powered agents can be accurate on average yet unreliable in production, a discrepancy that has been observed but remains largely unaddressed. When given the same task five times, a ReAct agent on the AppWorld benchmark using GPT-4.1 succeeds in all five runs only 53% of the time, even though its per-run pass rate averages 77%. We call this 24-point shortfall the consistency gap, and we argue that addressing it is a precondition for trustworthy AI agent deployment. We present a self-evolving agent framework that reduces this gap by identifying unstable, low-consistency steps in agent trajectories and converting them into episodic memory the agent can draw on in future runs. At its core is a Consistency Analyzer that pinpoints where and why a trajectory is likely to flip across executions, and a Guideline Generator that converts the diagnosis into targeted guidelines, committed to memory and injected into future agent executions on similar tasks. On AppWorld with ReAct/GPT-4.1, our framework raises the fraction of tasks that succeed in all five runs by +16 points on same-task evaluation and +13 points on similar-task generalization.
Evelyn Duesterwald, Benjamin Elder, Lilian Ngweta et al.· 0 citations
This work introduces TriArchitect, a novel shared-state multi-agent framework designed to bridge the gap between probabilistic reasoning and deterministic structural integrity, and introduces a rigorous Validator-Veto Protocol, which enforces a "compilation-as-consensus" rule to reject hallucinated proposals before they corrupt the codebase.
Neeraj Kumar Singh Beshane· International journal of com...· 0 citations
We use cookies to run the site and, with your consent, for analytics and to show ads.
See our Cookie Policy.
