APCD: Adaptive Path-Contrastive Decoding for Reliable Large Language Model Generation

arXiv:2605.09492v2 Announce Type: replace-cross
Abstract: Large language models (LLMs) often suffer from hallucinations due to error accumulation in autoregressive decoding, where suboptimal early token choices misguide subsequent generation. Although multi-path decoding can improve robustness by exploring alternative trajectories, existing methods lack principled strategies for determining when to branch and how to regulate inter-path interactions. We propose Adaptive Path-Contrastive Decoding (APCD), a multi-path decoding framework that improves output reliability through adaptive exploration and controlled path interaction. APCD consists of two components: (1) Entropy-Driven Path Expansion, which delays branching until predictive uncertainty – measured by Shannon entropy over top candidate tokens – indicates multiple plausible continuations; and (2) Divergence-Aware Path Contrast, which encourages diverse reasoning trajectories while dynamically attenuating inter-path influence as prediction distributions diverge. Experiments on eight benchmarks demonstrate improved factual accuracy while maintaining decoding efficiency. Our code is available at https://github.com/zty-king/APCD.