Code Generation by Differential Test Time Scaling

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: He, Yifeng, Wang, Ethan, Wang, Jicheng, Ouyang, Xuanxin, Chen, Hao
Format: Preprint
Published: 2026
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866918512956538880
author He, Yifeng
Wang, Ethan
Wang, Jicheng
Ouyang, Xuanxin
Chen, Hao
author_facet He, Yifeng
Wang, Ethan
Wang, Jicheng
Ouyang, Xuanxin
Chen, Hao
contents Test-time scaling has emerged as a promising approach for improving code generation by exploring large solution spaces at inference time. However, existing methods often rely on public test cases that are unavailable in practice, or require extensive LLM inference for candidate selection, leading to significant token consumption and time overhead. We present DiffCodeGen, a novel test-time scaling method for code generation based on coverage-guided differential analysis. DiffCodeGen generates diverse code candidates using various sampling and prompting strategies, then applies coverage-guided fuzzing to synthesize inputs without requiring any existing tests or large language models. By executing all candidates on these inputs, DiffCodeGen captures their dynamic behavior and clusters candidates based on behavioral similarity. DiffCodeGen selects the medoid of the largest cluster as the final output. Unlike prior test-time scaling methods that invoke additional LLM inference for candidate selection, DiffCodeGen performs selection without any extra model calls, incurring little to no additional token consumption. DiffCodeGen is fully asynchronous, naturally suited to the current trend of agentic coding, and is thus efficient and highly scalable. We evaluate DiffCodeGen across 4 large language models, demonstrating consistent improvements over baselines. Compared to state-of-the-art test-time scaling methods, DiffCodeGen achieves competitive or superior performance while using only a fraction of time and tokens. DiffCodeGen is model-agnostic and can be combined with reasoning models to further boost performance.
format Preprint
id arxiv_https___arxiv_org_abs_2605_20473
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle Code Generation by Differential Test Time Scaling
He, Yifeng
Wang, Ethan
Wang, Jicheng
Ouyang, Xuanxin
Chen, Hao
Software Engineering
Artificial Intelligence
Machine Learning
Test-time scaling has emerged as a promising approach for improving code generation by exploring large solution spaces at inference time. However, existing methods often rely on public test cases that are unavailable in practice, or require extensive LLM inference for candidate selection, leading to significant token consumption and time overhead. We present DiffCodeGen, a novel test-time scaling method for code generation based on coverage-guided differential analysis. DiffCodeGen generates diverse code candidates using various sampling and prompting strategies, then applies coverage-guided fuzzing to synthesize inputs without requiring any existing tests or large language models. By executing all candidates on these inputs, DiffCodeGen captures their dynamic behavior and clusters candidates based on behavioral similarity. DiffCodeGen selects the medoid of the largest cluster as the final output. Unlike prior test-time scaling methods that invoke additional LLM inference for candidate selection, DiffCodeGen performs selection without any extra model calls, incurring little to no additional token consumption. DiffCodeGen is fully asynchronous, naturally suited to the current trend of agentic coding, and is thus efficient and highly scalable. We evaluate DiffCodeGen across 4 large language models, demonstrating consistent improvements over baselines. Compared to state-of-the-art test-time scaling methods, DiffCodeGen achieves competitive or superior performance while using only a fraction of time and tokens. DiffCodeGen is model-agnostic and can be combined with reasoning models to further boost performance.
title Code Generation by Differential Test Time Scaling
topic Software Engineering
Artificial Intelligence
Machine Learning
url https://arxiv.org/abs/2605.20473