Rethinking in Spikes: Mitigating Hallucinations in MDLMs with Step-Aware Decoding

Zhongxing Xu, Zhonghua Wang, Zhe Qian, Shiyan Su, Ming Hu, Xiaocheng Zou, Wei Feng, Mingquan Lin, Yuyin Zhou, Yifan Peng, Hamid Rezatofighi, Zongyuan Ge

Rethinking in Spikes: Mitigating Hallucinations in MDLMs with Step-Aware Decoding: 12 upvotes on Hugging Face Daily Papers, #13 of 40 papers on 2026-11-16. Day-by-day upvote history. It lost 24 votes when the Hub removed votes in bulk.

Upvotes over time

It had 35 upvotes by its first day, 36 upvotes by the next day, 12 upvotes by day 3, 12 upvotes by day 7, and has 12 now. It ranked #43 of 146 papers in its week and #211 of 520 in its month. Its code repository has 23 GitHub stars.

Recent advancements in multimodal diffusion language models (MDLMs) have exhibited strong global modeling capabilities through iterative refinement. We observe that mask uncertainty follows an overall downward trend during iterative denoising, while a few denoising steps exhibit localized entropy spikes closely associated with hallucinations. We argue that anomalous steps can be identified by measuring the entropy deviation of each denoising step from its neighboring temporal window. We hypothesize that discrete token commitment discards the distributional information at masked positions, limiting the continued interaction between candidate semantics and the context during entropy spikes. With this goal, we present StepRefine, a plug-and-play decoding strategy that leverages semantic context to achieve reliable denoising. The core of our method lies in a step-aware switch between discrete denoising and latent refinement. The model employs probability-weighted embeddings at entropy spike steps to preserve diverse reasoning hypotheses and performs multiple rounds of continuous refinement. Meanwhile, a refinement controller monitors distributional stability to reduce unnecessary loops and switch the model back to standard discrete denoising. Through extensive experiments, StepRefine demonstrates significant hallucination mitigation across different MDLMs on multimodal benchmarks.

Paper page on Hugging Face · arXiv

Data: hysts-bot-data/daily-papers-stats and the Daily Papers API. Open data: tardellirs/paper-pulse-data. Sister project: Model Pulse, the download history of every model on the Hub.