iterative refinement

In industrial quality inspection, anomaly detection suffers from poor robustness due to high noise levels and sparse defective samples. To address this, we propose Iterative Refinement of Pseudo-labels (IRP), a self-supervised method that alternately evaluates sample credibility and removes misleading instances under feature-space consistency constraints—effectively purifying the training set dynamically without human annotations and generating high-fidelity self-supervised signals. IRP introduces the novel paradigm of “iterative data refinement,” significantly enhancing model robustness against label noise and cross-domain generalization capability. Evaluated on KSDD2 and MVTec AD benchmarks, IRP consistently outperforms existing unsupervised and self-supervised methods. Notably, under high-noise conditions, it achieves substantial improvements in detection accuracy and reduces false positive rates by over 25%.

This study addresses the common omission in current large language model evaluations of code generation—the iterative refinement process inherent in real-world programming and the models’ capacity for self-correction using feedback. The authors propose a novel framework that leverages execution-based feedback, such as compilation errors and test failures, to systematically investigate how reasoning and non-reasoning models utilize such signals across multiple programming languages. Through multidimensional categorization of code failures and extensive cross-model, cross-language experiments, they demonstrate that reasoning models consistently improve over iterations and significantly outperform non-reasoning counterparts. While syntactic and runtime errors prove relatively amenable to correction, logical and algorithmic errors remain challenging, thereby delineating the current limits of feedback-driven repair mechanisms.

Traditional evaluation of code generation and repair relies solely on binary correctness, overlooking critical dimensions such as incremental improvement, feedback utilization, and behavioral preservation during the repair process. To address this limitation, this work proposes PAIR-Bench, a novel benchmark that guides large language models through feedback-driven iterative repair via a structured feedback protocol combining failure-region control and prompt-depth control. The benchmark introduces progressive prompting, grouping of failure scenarios, and multi-level repair guidance, alongside trajectory-level evaluation metrics. For the first time, it enables fine-grained, adaptive assessment of the code improvement process, outperforming conventional coarse-grained evaluation paradigms in terms of target repair efficacy, generalization capability, behavioral consistency, and required human assistance.

Existing LLM-based automated program repair (APR) methods rely on coarse-grained test feedback and lack fine-grained error attribution, resulting in low repair efficiency and severe error propagation. To address this, we propose TokenRepair—the first dual-layer APR framework integrating internal reflection and external feedback. It first identifies potentially erroneous tokens via context-aware, token-level uncertainty analysis; then performs precise patch generation through chain-of-thought-guided local rewriting and quality-aware iterative optimization—avoiding global regeneration to significantly improve repair stability and accuracy. Evaluated on Defects4J 1.2 and HumanEval-Java, TokenRepair successfully repairs 88 and 139 bugs, respectively, outperforming state-of-the-art methods by 8.2%–34.9%. These results empirically validate the effectiveness and generalizability of our token-level fine-grained reflection mechanism.

Automatic Program Repair (APR) faces stringent constraints—specifically, generating at most ten patches per single defect—necessitating highly efficient and precise patch generation. Method: We propose a test-feedback-driven self-iterative patch generation framework, systematically evaluating full-parameter fine-tuning versus LoRA adaptation across three instruction-tuned LLMs (DeepSeekCoder-Instruct, CodeLlama-Instruct, Llama3.1-Instruct) on datasets of 1K, 30K, and 65K samples. Contribution/Results: Our empirical study is the first to demonstrate that fine-tuning less than 1% of training data improves plausible patch rate by 78%. Iterative refinement significantly benefits base models and remains indispensable for fine-tuned models on complex defects. We identify a diminishing-return inflection point in fine-tuning: excessive adaptation induces overfitting and degrades performance. On HumanEval-Java and Defects4J, our approach substantially improves repair rates, challenging the prevailing claim that full-parameter fine-tuning is ineffective for APR.

This work addresses the challenge of repeatedly verifying the safety of cyber-physical systems during iterative design, where frequent reconfigurations necessitate costly global revalidation. To overcome this, the paper introduces the principle of “refactoring-as-proposition,” which, for the first time, formalizes hybrid system refactoring as provable logical propositions. Leveraging differential refinement logic (dRL), the approach uniformly characterizes system properties and their preservation across refactorings—including those involving auxiliary variables—and enables localized verification. By transferring safety proofs from the original system to its refactored variant, the method substantially reduces verification complexity, facilitates automated or modular proof construction, and eliminates the need for exhaustive revalidation of the entire system.

This study addresses the lack of systematic understanding regarding the impact of repair loop iteration counts in large language model (LLM)-based software engineering tasks, where prior work often relies on arbitrarily defined repair budgets. Through a cross-task (code generation, test generation, code translation) and cross-model empirical analysis, this work reveals—for the first time—a pronounced diminishing marginal returns phenomenon in iterative repair: performance gains are concentrated within the first 3–4 iterations, with negligible improvements thereafter. The findings underscore that the design of the repair workflow and feedback mechanisms exerts a far greater influence on repair efficacy than the choice of LLM itself. The authors advocate for treating repair budget as a critical experimental variable to ensure reliable, computationally efficient, and reproducible evaluation outcomes.

This work addresses the lack of auditability and reproducibility in existing autonomous improvement loops, where failures often go unrecorded and repair decisions are decoupled from agent history. To remedy this, the authors propose ActiveGraph, an event-sourcing–based runtime that embeds the entire improvement process within the agent’s historical trace, ensuring end-to-end auditability. The approach incorporates a retention-set gating mechanism, failure-mode–aware routing, and prompt probes to diagnose failures, generate repairs in typed pipeline nodes, and validate them through static checks, sandboxed execution, in-sample evaluation, and retention-set verification before promotion. Evaluated on LongMemEval-S, the method yields significant accuracy gains in four out of five retention partitions (+0.05 to +0.10), with a modest +0.01 improvement in the fifth, indicating that the primary bottleneck lies in evidence integration rather than retrieval.

This work addresses the degradation of repair correctness in existing code repair agents, which often lose valid patches during repetitive generate–test–revise cycles. The study introduces a novel decomposition of the repair process into five orthogonal dimensions: admission, retention, certification, capability, and liveness. It proposes an executable specification mechanism grounded in state-bound evidence and typed revision contracts, along with an auditable reference implementation. Through branch-freezing for bias control, execution trajectory tracking, and formal verification, experiments reveal that stale execution trajectories significantly reduce repair success rates. The proposed approach effectively mitigates this degradation; however, it does not enhance overall repair capability, thereby highlighting the critical influence of system component heterogeneity and underlying bottlenecks.

Current research on AI self-improvement lacks a systematic distinction between types of improvement and the degree of human-AI closed-loop interaction, impeding a clear understanding of the boundaries and risks of recursive self-improvement. This work addresses this gap by analyzing 1,250 arXiv papers and proposing the first dual-axis taxonomy that integrates improvement objectives—encompassing deployment behavior, policy training, evaluator optimization, and scientific research workflows—with levels of closed-loop autonomy. The study highlights the central role of self-evaluation signals within validation hierarchies and demonstrates that the intensity of self-improvement critically depends on the validation level at which these signals operate. It identifies “research direction setting” as a key bottleneck requiring human intervention and underscores governance-level metrics for self-improvement as the most underexplored area in current scholarship.