Eliminating Hallucination-Induced Errors in LLM Code Generation with Functional Clustering

TL;DR

This paper introduces functional clustering, a black-box method that significantly reduces hallucination-induced errors in LLM code generation by clustering candidate outputs based on I/O behavior and using empirical confidence estimates.

Contribution

The paper presents a novel functional clustering wrapper that eliminates nearly all hallucination errors in LLM code generation while providing a tunable confidence score, applicable to closed-source models.

Findings

Reduces error rate from ~65% to 2% on LiveCodeBench

Preserves baseline pass@1 on solvable tasks

Achieves 0% error at conservative thresholds while answering 15.6% of prompts

Abstract

Modern code-generation LLMs can already solve a large fraction of programming problems, yet they still hallucinate subtle bugs that make their outputs unsafe for autonomous deployment. We present functional clustering, a black-box wrapper that eliminates nearly all hallucination-induced errors while providing a tunable confidence score. The wrapper samples many candidate programs, executes each on a self-generated test suite, and clusters candidates whose I/O behavior is identical; the empirical mass of the largest cluster serves as an exact confidence estimate. A single scalar threshold on this estimate lets users trade coverage for reliability with exponential guarantees. On LiveCodeBench our verifier preserves baseline pass@1 on solvable tasks yet slashes the error rate of returned answers from ~65% to 2%, and drives it to 0% at a conservative threshold while still answering 15.6% of prompts. Manual audits show that the few residual mistakes stem from prompt misinterpretation, not random generation noise, narrowing future work to specification clarity. Because the method requires only sampling and sandbox execution, it applies unchanged to closed-source APIs and future models, offering a practical path toward dependable, autonomous code generation. Our code is available on Github (https://github.com/20ChaituR/functional-clustering).