Reproduce “Compare to Correct: Pairwise-Guided Logit Adjustment for LLM Test-Time Scaling” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Cong Hua, Qianqian Xu, Zitai Wang, Zhiyong Yang, Shilong Bao, Linchao Zhang, Qingming Huang
The authors propose PGLA, which adjusts voting among sampled language-model solutions using confidence comparisons between paths supporting different answers. These comparisons serve as a proxy for preferences, not judgments of correctness. On AIME 2024, Qwen3-8B in reasoning mode with 32 responses scored 87.00% accuracy, versus 81.67% for DeepConf-Weighted SC, within the evaluated settings.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.