Despite their impressive capacities, Large language models (LLMs) often struggle with the hallucination issue of generating inaccurate or fabricated content even when they possess correct knowledge. In this paper, we extend the exploration of the correlation between hidden-state prediction changes and output factuality into a deeper, token-wise level. Based on the insights , we propose cross-layer Entropy eNhanced Decoding (END), a decoding method that mitigates hallucinations without requiring extra training. END leverages inner probability changes across layers to individually quantify the factual knowledge required for each candidate token, and adjusts the final predicting distribution to prioritize tokens with higher factuality. Experiments on both hallucination and QA benchmarks demonstrate that END significantly enhances the truthfulness and informativeness of generated content while maintaining robust QA accuracy. Moreover, our work provides a deeper perspective on understanding the correlations between inherent knowledge and output factuality.
Integrative Decoding: Improve Factuality via Implicit Self-consistency
The Hong Kong Polytechnic University（香港理工大学）；Tsinghua University（清华大学）；Microsoft Research（微软研究院）；Microsoft Azure AI（微软Azure人工智能）；University of Illinois at Urbana-Champaign（伊利诺伊大学厄巴纳-香槟分校）
DHI: Leveraging Diverse Hallucination Induction for Enhanced Contrastive Factuality Control in Large Language Models
School of Computer Science, Wuhan University（武汉大学计算机学院）；Tsinghua Shenzhen International Graduate School, Tsinghua University（清华大学深圳国际研究生院）；School of Artificial Intelligence, Wuhan University（武汉大学人工智能学院）
Invisible Ink, Visible Lies: How Production Watermarking Causes LLMs to Hallucinate
Nanjing University of Science and Technology（南京理工大学）；Anhui University of Technology（安徽工业大学）；The Hong Kong Polytechnic University（香港理工大学）；University of Luxembourg（卢森堡大学）；CSIRO（澳大利亚联邦科学与工业研究组织）
Sapienza University of Rome（罗马第一大学）；Facebook AI（脸书人工智能研究院）；University College London（伦敦大学学院）
Recent progress in pre-trained language models led to systems that are able to generate text of an increasingly high quality. While several works have investigated the fluency and grammatical correctness of such models, it is still unclear to which extent the generated text is consistent with factual world knowledge. Here, we go beyond fluency and also investigate the verifiability of text generated by state-of-the-art pre-trained language models. A generated sentence is verifiable if it can be corroborated or disproved by Wikipedia, and we find that the verifiability of generated text strongl…
HalluTracer: Pre-Decoding Truthfulness Prediction via Depth-Averaged Probe-Logit
University of Technology Sydney（悉尼科技大学）；City University of Macau（澳门城市大学）；Meta；actAVA AI（actAVA人工智能公司）；Microsoft AI（微软人工智能）；Squirrel Ai Learning（松鼠人工智能学习公司）；East China Normal University（华东师范大学）
SpecBound: Adaptive Bounded Self-Speculation with Layer-wise Confidence Calibration
Key Laboratory of Intelligent Information Processing, Institute of Computing Technology, Chinese Academy of Sciences（智能信息处理重点实验室，计算技术研究所，中国科学院）；State Key Laboratory of AI Safety, Institute of Computing Technology, Chinese Academy of Sciences（人工智能安全国家重点实验室，计算技术研究所，中国科学院）；University of Chinese Academy of Sciences, Beijing, China（中国科学院大学，北京，中国）
Escaping Confidence Trap: Evolutionary Decoding for Mathematical Reasoning in Diffusion LLMs
Australian National University（澳大利亚国立大学）；Nanyang Technological University（南洋理工大学）；National University of Singapore（新加坡国立大学）；Amazon（亚马逊）；University of Technology Sydney（悉尼科技大学）
ReasonAlloc: Hierarchical Decoding-Time KV Cache Budget Allocation for Reasoning Models
Tsinghua University（清华大学）；City University of Hong Kong（香港城市大学）；Peking University（北京大学）；Shenzhen University of Advanced Technology（深圳理工大学）；Institute of Automation, Chinese Academy of Sciences（中国科学院自动化研究所）
Technical University of Darmstadt（达姆施塔特工业大学）；RWTH Aachen University（亚琛工业大学）；AppTek GmbH（AppTek有限公司）
In this work, we present a model for document-grounded response generation in dialog that is decomposed into two components according to Bayes theorem. One component is a traditional ungrounded response generation model and the other component models the reconstruction of the grounding document based on the dialog context and generated response. We propose different approximate decoding schemes and evaluate our approach on multiple open-domain and task-oriented document-grounded dialog datasets. Our experiments show that the model is more factual in terms of automatic factuality metrics than t…
