Lost in Diffusion: Uncovering Hallucination Patterns and Failure Modes in Diffusion Large Language Models

TL;DR

This paper investigates hallucination patterns in diffusion large language models, revealing they are more prone to hallucination than autoregressive models and identifying unique failure modes affecting reliability.

Contribution

It provides the first controlled comparative analysis of hallucination in dLLMs, highlighting their distinct failure modes and divergence in inference dynamics.

Findings

dLLMs exhibit higher hallucination propensity than AR models.

Non-sequential decoding enables potential for continuous refinement.

Identified failure modes include premature termination and context intrusion.

Abstract

While Diffusion Large Language Models (dLLMs) have emerged as a promising non-autoregressive paradigm comparable to autoregressive (AR) models, their faithfulness, specifically regarding hallucination, remains largely underexplored. To bridge this gap, we present the first controlled comparative study to evaluate hallucination patterns in dLLMs. Our results demonstrate that current dLLMs exhibit a higher propensity for hallucination than AR counterparts controlled for architecture, scale, and pre-training weights. Furthermore, an analysis of inference-time compute reveals divergent dynamics: while quasi-autoregressive generation suffers from early saturation, non-sequential decoding unlocks potential for continuous refinement. Finally, we identify distinct failure modes unique to the diffusion process, including premature termination, incomplete denoising, and context intrusion. Our findings underscore that although dLLMs have narrowed the performance gap on general tasks, their distinct hallucination mechanisms pose a critical challenge to model reliability. Our code is available at https://github.com/ZeroLoss-Lab/Lost-in-Diffusion