Jun 2026· arXiv.org· Vol abs/2606.30810· 0 citations· 32 references
Computer Science
TL;DR
CURE is developed, a contrastive unlearning approach that shifts unlearning from purely suppressing outdated knowledge to explicitly promoting correct API replacements, enabling more reliable adaptation to evolving software libraries.
Abstract
Large Language Models (LLMs) have recently achieved strong performance in code generation. However, due to knowledge cut-off and the rapid evolution of software libraries, they often generate deprecated API usages that lead to unreliable and incompatible code. Existing fine-tuning methods lack selectivity when only a small portion of model knowledge requires modification. Recent model-level approaches, such as machine unlearning and model editing, offer a promising direction for modifying parametric knowledge. However, their use for deprecated API mitigation remains largely unexplored. Moreover, existing methods primarily suppress outdated APIs, but do not explicitly steer models toward correct replacements, often leading to mismatched or incomplete generations. To address this limitation, we developed CURE, a contrastive unlearning approach that shifts unlearning from purely suppressing outdated knowledge to explicitly promoting correct API replacements. Concretely, CURE jointly discourages deprecated APIs while encouraging their valid alternatives, enabling more reliable adaptation to evolving software libraries. The experiments on recent deprecated API benchmark dataset show that CURE not only reduces deprecated API usage but also improves correct API replacement, while preserving general code generation performance. CURE substantially outperforms two SOTA baselines with respect to different quality metrics. These findings highlight the importance of combining suppression with replacement when adapting LLMs to evolving software ecosystems.
HELP, a program optimizer that prompts an LLM to generate diverse, non-overlapping atomic edits and then systematically selects and composes them into coherent program improvements using evaluator scores, is introduced.
Jingwen Fu, Zhen Liu, Yuhan Liu et al.· 2 citations
Regular expressions (regexes) are widely used in software development but remain difficult to author and validate due to their compact syntax and subtle semantics. While large language models (LLMs) can now generate regexes from natural-language descriptions, their outputs often miss developer intent, and existing refinement techniques rely on developers to craft positive and negative examples—especially discriminatory ones that expose fine-grained semantic differences. Producing such examples is cognitively demanding and often leaves regexes under-validated. This paper introduces ExReg, a human-in-the-loop workflow that shifts this burden away from developers. Given an ambiguous natural-language specification, an LLM first proposes multiple plausible regex candidates. Instead of requiring developers to devise discriminative examples, the system automatically generates them using SMT-based constraint solving. Developers need only affirm whether these synthesized test strings match their intent. The system further mutates candidates and reuses discriminative strings to systematically eliminate incorrect patterns. Across six benchmark datasets and six state-of-the-art LLMs, ExReg accurately identifies valid regexes—or determines that none are appropriate—while requiring only minimal developer validation. In particular, ExReg achieves an average accuracy of 87%, while requiring users to inspect only 5 examples on average, each with a mean length of 8 characters and an inter-example waiting time of under 13 seconds. These results highlight how automated example generation guided by formal methods and mutations can improve the reliability, efficiency, and trustworthiness of LLM-assisted regex pattern generation.
Hadar Muchtar, Ilia Shevrin, S. Maoz· ACM Transactions on Software...· 0 citations
Large Language Models (LLMs) are predominantly assessed based on their common sense reasoning, language comprehension, and logical reasoning abilities. While models trained in specialized domains like mathematics or coding have demonstrated remarkable advancements in logical reasoning, there remains a significant gap in evaluating their code generation capabilities. Existing benchmark datasets fall short in pinpointing specific strengths and weaknesses, impeding targeted enhancements in models’ reasoning abilities to synthesize code.
To bridge this gap, this thesis introduces two novel contributions: CodeEval and CodeQual. CodeEval is an innovative, pedagogical benchmarking method that mirrors the evaluation processes encountered in academic programming courses. It comprises a multi-dimensional benchmark dataset of 602 hand-crafted problems designed to rigorously evaluate LLMs across 24 distinct aspects of Python programming, covering three proficiency levels—beginner, intermediate, and advanced—and includes both class-based and function-based problem types with detailed problem specifications and comprehensive test suites achieving 99.1% coverage. To facilitate widespread adoption, we developed RunCodeEval, an open-source execution framework that provides researchers with a ready-to-use evaluation pipeline. Our evaluation of 15 state-of-the-art LLMs revealed consistent performance degradation with increasing complexity (validated statistically, Cohen’s d = 0.790) and universal struggles with advanced concepts like concurrency.
Code quality is inherently subjective, encompassing dimensions like readability, efficiency, and adherence to language idioms that traditional static metrics fail to capture adequately. While large language models can assess these subjective qualities, lightweight models offer practical advantages: seamless CI/CD pipeline integration, lower operational costs, and full control over model behavior. We investigate whether such models can learn to assess code quality by training on synthetic LLM annotations. We introduce CodeQual, a dataset of 5,819 code samples derived from five established sources spanning diverse domains—competitive programming, pedagogical problems, software engineering, and general benchmarks—scored by LLMs across five quality dimensions, with 655 human-annotated samples for evaluation. Our fine-tuned model, CodeQualBERT, not only matches LLM performance but exceeds inter-human agreement on all five dimensions, achieving 16–100% improvement over the inter-human agreement baseline.
Together, these contributions provide a comprehensive framework for evaluating and improving LLMs in software engineering contexts, encompassing both functional correctness assessment and subjective code quality evaluation.
Large Language Models (LLMs) increasingly require selective removal of harmful or sensitive knowledge, called unlearning, yet existing methods and benchmarks fail to evaluate this capability completely. Current approaches rely on disjoint forget and retain sets composed of independent facts, and measure success using simple and direct factual recall. This framing fails to capture a key requirement of unlearning, namely the ability to eliminate harmful behaviors while preserving benign and beneficial knowledge. We argue that effective unlearning must operate at the level of concepts, ensuring complete removal of unsafe applications while maintaining their correct and useful usage, thereby achieving conceptually meaningful and complete unlearning. To better evaluate unlearning techniques from such a practical viewpoint, we introduce the notion of dual-use concepts: concepts that can be used in both harmful and benign contexts. Building on these concepts, we construct a benchmark called ConceptGuard where forget and retain sets are explicitly complementary in concept usage. Our benchmark uniquely enables unlearning to be explored and gauged at the level of concepts, instead of sparse facts, and evaluation is intent-sensitive with the goal of maximizing contextual separation to promote safer behavior. We demonstrate that current unlearning techniques perform poorly under this setting, showing weak contextual separation alongside poor performance in ROUGE and concept-level metrics. Our results reveal strong forgetting-utility trade-offs, limited gains in contextual sensitivity, and poor consistency in concept-level control across methods, and provide ideas for unlearning approaches that better align with real-world safety requirements. Our dataset is publicly available.
SCOPE is presented, a prover-initialized subgoal critic for code generation that adapts a Lean-oriented prover model to produce three parseable feedback fields for downstream code generation: subgoals, gap analysis, and a robustness checklist.
Yueke Zhang, Yifan Zhang, Zihan Fang et al.· 0 citations
Natural-language requirements for program synthesis are often incomplete or ambiguous, yet large language models are commonly expected to generate code in a single pass. Prior clarification-based methods address this issue by asking follow-up questions when sampled candidate programs disagree, but fixed clarify-on-disagreement policies can overuse clarification and can also overtrust weak behavioral agreement. We present an adaptive routing framework for LLM-based program synthesis that treats clarification as an inference-time control decision. The framework augments a ClarifyGPT-style pipeline with execution-driven confidence estimation, semanticdifference analysis, and bounded candidate expansion, allowing the system to choose among direct generation, additional evidence gathering, and clarification. We evaluate the framework on MBPP, HumanEval, and extended-test variants using GPT-4.1 mini, Claude Haiku 4.5, and GPT-5.4 mini. Adaptive routing improves pass@1 accuracy by up to 7.60 percentage points over single-pass baselines. Compared with fixed-policy clarification, it preserves accuracy while reducing token usage by up to 57.2% for GPT-4.1 mini, and reallocates computation toward harder cases for Claude Haiku 4.5. These results suggest that clarification is most useful when triggered selectively based on execution evidence and semantic disagreement, even when ambiguity is observed indirectly through candidate behavior rather than through explicitly annotated ambiguous requirements.
Muhammad Ahmed, Edwar Tiu, Niyati Nikunj Kapadia et al.· Annual International Compute...· 0 citations