mlsanityData to 7 Oct 2026

Reliability Challenges in Diffusion Vision-Language Models

arXiv 2609.01318 ↗First posted 1 Sep 2026cs.CV cs.CLcited by 0

On the map: Multimodal LLMs / vision-language models → Image editing and quality assessment

also in 1 more
Abstract (arXiv)

Diffusion-based Large Vision-Language Models (dLVLMs) have recently emerged as a compelling alternative to autoregressive (AR) LVLMs, offering advantages in parallel decoding, bidirectional context, and controllable generation. Despite rapid progress, their reliability properties remain largely uncharacterized. We present the first systematic reliability evaluation of hallucination and bias in dLVLMs, benchmarking six diffusion models against competitive AR baselines across four dimensions. Our key findings are: (1) dLVLMs reverse the yes-bias of AR models in binary visual queries; (2) they achieve competitive hallucination rates yet exhibit degraded linguistic quality; (3) they collapse to near-zero accuracy on underrepresented racial groups with opposite-polarity gender bias; and (4) they exhibit accuracy collapse in multiple-choice settings when the correct option is shorter than its distractors, associated with a length prior that emerges at the first denoising step. Tokens committed at late denoising steps with low confidence further correlate with hallucinated content, pointing to a mechanistic signal unique to diffusion generation. These patterns vary across model families, suggesting reliability is shaped by the generative paradigm together with training data.

Key facts. No citing papers in the map yet (data through 7 Oct 2026). Placed in Image editing and quality assessment, one of 16 directions in Multimodal LLMs / vision-language models. Builds on 8 of its 32 references that are in the map. First posted on arXiv 1 Sep 2026.

Builds on

8 of 32 references.

Closest ancestors · 1

Same direction, newest first

  1. 2025Dimple: Discrete Diffusion Multimodal Large Language Model with Parallel Decoding22 May

Other references · 7

  1. 2025Qwen2.5-VL Technical Reportcited by 5,418
  2. 2024GPT-4o System Cardcited by 4,467
  3. 2024Expanding Performance Boundaries of Open-Source Multimodal Models with Model, Data, and Test-Time Scalingcited by 1,573
  4. 2023mPLUG-Owl: Modularization Empowers Large Language Models with Multimodalitycited by 1,038
  5. 2025Dream 7B: Diffusion Large Language Modelscited by 515
  6. 2025LLaDA-MoE: A Sparse MoE Diffusion Language Modelcited by 54
  7. 2025Dream-VL & Dream-VLA: Open Vision-Language and Vision-Language-Action Models with Diffusion Language Model Backbonecited by 18

Built on by

Not cited yet.

Similar, not cited

Close in wording, no citation either way

  1. 2026Lost in Diffusion: Uncovering Hallucination Patterns and Failure Modes in Diffusion Large Language Models
  2. 2025DiffusionVL: Translating Any Autoregressive Models into Diffusion Vision Language Models
  3. 2024DEEM: Diffusion Models Serve as the Eyes of Large Language Models for Image Perception
  4. 2026Before the Token Commits: Trajectory-Level Benchmarking of Visual Hallucinations in Diffusion VLMs
  5. 2024VALOR-EVAL: Holistic Coverage and Faithfulness Evaluation of Large Vision-Language Models
  6. 2025Curing Semantic Drift: A Dynamic Approach to Grounding Generation in Large Vision-Language Models
  7. 2024Simple and Effective Masked Diffusion Language Models
  8. 2025From Denoising to Refining: A Corrective Framework for Vision-Language Diffusion Model