APCD: Adaptive Path-Contrastive Decoding for Reliable Large Language Model Generation
Fuente:
arXiv
Guardado en:
| Autores principales: | , , |
|---|---|
| Formato: | Preprint |
| Publicado: |
2026
|
| Materias: | |
| Acceso en línea: | |
| Etiquetas: |
Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
|
| _version_ | 1866913149222912000 |
|---|---|
| author | Zheng, Tianyu Wu, Hong Zhong, Jiaji |
| author_facet | Zheng, Tianyu Wu, Hong Zhong, Jiaji |
| contents | 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. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2605_09492 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | APCD: Adaptive Path-Contrastive Decoding for Reliable Large Language Model Generation Zheng, Tianyu Wu, Hong Zhong, Jiaji Computation and Language Artificial Intelligence 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. |
| title | APCD: Adaptive Path-Contrastive Decoding for Reliable Large Language Model Generation |
| topic | Computation and Language Artificial Intelligence |
| url | https://arxiv.org/abs/2605.09492 |