It is suggested that cheap uncertainty estimators are insufficient on their own to improve code correctness, and that their practical value lies in serving as gating signals for costlier execution-based correction loops rather than as standalone substitutes for verification.
Abstract
Large language models for code generation often produce incorrect solutions without reliable indicators of failure. We study whether uncertainty estimation methods developed for natural language transfer to code generation, and whether such signals can improve code generation via selective self-correction. We evaluate five uncertainty methods: mean token entropy, verbalized confidence, $P(\text{True})$, entropy ensembles, and semantic entropy probes, across three small code LLMs on HumanEval and BigCodeBench. We find that multi-sample $P(\text{True})$ achieves the strongest correlation with correctness, while all the other methods, including semantic entropy probes, yield only weak correlation. We then use these uncertainty signals to drive three self-correction policies: adaptive decoding, uncertainty-based regeneration, and verification-based regeneration. Our results reveal a stronger negative finding than anticipated: uncertainty-based self-correction fails to reliably improve Pass@1, degrading accuracy in 5 of 6 configurations across both benchmarks ($-3$pp to $-10$pp), and adaptive decoding degrades accuracy in 4 of 6 configurations. Only verification-based self-correction reliably improves Pass@1, with gains of $+6$ to $+26$ percentage points on HumanEval and $+8$ to $+20$ percentage points on BigCodeBench, scaling inversely with baseline strength. These findings replicate consistently across both benchmarks and suggest that cheap uncertainty estimators are insufficient on their own to improve code correctness, and that their practical value lies in serving as gating signals for costlier execution-based correction loops rather than as standalone substitutes for verification.
Prediction uncertainty is a widely adopted metric for quantifying model confidence, with downstream applications spanning model explanation, data selection, and prediction rollback. Despite its demonstrated utility, the potential of uncertainty quantification to enhance code generation in large language models (LLMs) remains largely underexplored, raising a critical question: to what extent can uncertainty serve as an effective signal for improving LLM-based code generation? To answer this question, we study uncertainty-aware rollback decoding, an inference-time strategy that uses uncertainty signals to identify unreliable generation regions and roll back to earlier valid prefixes without retraining the model. We evaluate this framework on seven code LLMs, five code generation benchmarks, and eight token-level uncertainty signals under a unified decoding setup. Our results show that the complete rollback framework improves over equal-budget restart across the evaluated benchmarks and model settings, with gains of up to 0.26 in pass@1 and 0.35 in AvgTestPassRate on functional code generation benchmarks, and an absolute improvement of up to 6.4\% in Patch-Aligned Safe Rate on Dsec-Python. Among the evaluated signals, information-theoretic measures such as token entropy and negative log-likelihood show the most favorable overall trend, frequently achieving the best or near-best results on standard benchmarks. A component-controlled ablation further shows that feedback-guided rollback provides the main improvement, while uncertainty localization provides an additional gain when checking, budget, rollback, and branch decay are held fixed.
Xianzong Wu, Xiaohong Li, Yuejun Guo et al.· 0 citations
Large language models (LLMs) have made notable progress in code generation, but they still struggle on challenging tasks that require sophisticated algorithms or complex implementations. Recent methods increasingly use code execution as feedback, especially in self-correction pipelines that construct verification signals from generated code. However, these pipelines often depend on supervision signals whose reliability is unknown, which can introduce misleading feedback, unnecessary revisions, and incorrect final answers. To address this issue, we propose ExeCRE, an Execution-Consistency guided code Reliability Estimation framework. Instead of judging candidate code by tests or LLM feedback, ExeCRE estimates code reliability by statistically analyzing consistency patterns in execution outputs over a large number of randomly generated inputs. It collects execution outputs over generated inputs, projects them into consistency signals, and applies the Dawid-Skene model to infer latent code reliability. We integrate ExeCRE into self-correction for code generation. Experiments show that ExeCRE consistently improves both effectiveness and stability, while substantially reducing misleading correction signals. Under GPT-5.2 on LiveCodeBench, the average number of misleading feedback cases on already correct code drops from 113.2 with a representative self-correction baseline to 14.0 with ExeCRE. As an additional study, we apply the same reliability estimation strategy to code-based mathematical reasoning and observe similar benefits. These results suggest that ExeCRE enables more reliable use of generated code in execution-based pipelines.
Yirui Dong, Richong Zhang, Fanshuang Kong et al.· 0 citations
Large language models (LLMs) are increasingly used as graders, verifiers, and process auditors, but most mathematical evaluations still emphasize final-answer accuracy. This can obscure whether a model can verify a non-canonical but valid solution trace. We introduce a controlled linear-equation benchmark for evaluating LLMs in the evaluator role. Each instance asks the model to judge final-answer correctness, step-level trace correctness, and the first incorrect step. Our evaluation of state-of-the-art open LLMs reveals a significant robustness gap: models that accurately evaluate canonical solutions often fail when presented with perturbed but logically equivalent variants. Across GPT-OSS 20B, Qwen3-14B, and Phi-4-Reasoning, base models perform well on canonical traces but degrade substantially on perturbed traces, especially for error localization. On valid perturbed traces, base-model false-rejection rates reach 75.6-85.3%, showing strong sensitivity to canonical solution form. Supervised fine-tuning, distillation, and test-time compute improve robustness in some settings, but gains are model dependent and can trade off against canonical performance. The results show that reliable process-level verification remains challenging, and evaluator robustness should be measured separately from solver accuracy, even in a simple algebraic domain with exact ground truth.
Code translation, as a challenging and fundamental task, is increasingly relying on large language models (LLMs). However, LLMs often give seemingly plausible but fallacious translations, misleading and even deceptive to debugging developers. We propose tHinter, an automated approach that frames translation error localization as a differential testing task. By fuzzing and a novel Bayesian-inspired algorithm that integrates developer heuristics into SBFL, tHinter thoroughly exposes translation errors and effectively pinpoints functionality-twisting errors. We evaluate tHinter in a Python to C++ scenario, a representative and challenging task for migrating high-level logic to performance-critical systems. Experiments show that tHinter achieves the highest F1 in localizing translation errors, outperforming all baselines. When debugging LLM-translated code with human efforts, it averagely reduces 71% lines developers need to review. Furthermore, we provide a comprehensive evaluation reference beyond traditional fault localization metrics. We integrate mixed-factorial user studies, expert validation, and SWOT-based strategic analysis to assess the perceived helpfulness and resilience within the rapidly evolving LLM landscape, offering a methodologically rigorous example for researchers in the context of AI-augmented software engineering.
Shengnan Wu, Xin-Yu Sun, Xin Wang et al.· ACM Transactions on Software...· 0 citations
It is observed that generated code often omits basic input validation or memory-safety checks, which can lead to overflows, resource exhaustion, or other reliability/security issues, and even the largest models frequently make simple mistakes.
Rodrigo Pato Nogueira, Marco Vieira, João R. Campos· 0 citations
CoGate is proposed, a confidence-gated co-decoding approach that controls the expert's influence on the co-decoding process based on its confidence, and outperforms existing co-decoding methods (CoSec+) across multiple benchmarks.
Minghao Hu, Lannan Luo, Allen G. Roush et al.· arXiv.org· 0 citations
We use cookies to run the site and, with your consent, for analytics and to show ads.
See our Cookie Policy.
