titleAVTR: Adaptive Multimodal Reasoning with Visual Imagination
authorsXiaoqian Shen, Yunyang Xiong, Changsheng Zhao, Saksham Suri, Chong Zhou, Zechun Liu, Zhipeng Cai, Zhuang Liu, Qi Qian, Wei Wen, Hongyu Xu, Hanxian Huang, Ernie Chang, Yanbei Chen, Vikas Chandra, Raghuraman Krishnamoorthi, Mohamed Elhoseiny, Lemeng Wu
date2026-11-18
venuearXiv (preprint, not peer reviewed)
idarXiv:2611.24161
affiliationMeta
upvotes8 (Hugging Face Daily Papers, 2026-11-21)
linksabs pdf hf code
cite
relatedSharpening Tax in Post-Training
World Action Modeling with Progressive Visual Planning
Native Action-Prior Learning from Videos for World Action Models
InteractiveWorld: A Verifiable Human-Centric Environment for Building Better Interactive MLLM Assistants
Context Language Models
abstractVision-Language Models increasingly rely on textual reasoning to solve advanced multimodal tasks. Despite its effectiveness, this paradigm faces two fundamental limitations. First, relying on textual reasoning alone can limit perceptual ability in vision-centric questions, where accurate answers depend on interpreting task-relevant visual evidence. Second, applying textual reasoning uniformly across inputs overlooks question-dependent variation in both reasoning modality and depth, often producing verbose and unnecessary traces when little logical inference is required. To this end, we propose Adaptive Visual-Textual Reasoning (AVTR), which introduces visual imagination as a complementary reasoning modality to textual thinking, enabling the model to revisit its visual observations and distill the salient evidence through generative visual reflection. Furthermore, we formulate reasoning-modality selection as a learned routing problem, allowing a single policy to adaptively harness the complementary strengths of visual reflection and textual reasoning in response to the perceptual and inferential demands of the input. Across 16 benchmarks, AVTR improves the average score by 4.5 points over Qwen3.5 baseline, while reducing inference latency by 33\%. It also scales effectively into a smaller size backbone, outperforming the evaluated 2B-scale baselines.

bibtex

@misc{shen2026avtr,
  title={AVTR: Adaptive Multimodal Reasoning with Visual Imagination},
  author={Xiaoqian Shen and Yunyang Xiong and Changsheng Zhao and Saksham Suri and Chong Zhou and Zechun Liu and Zhipeng Cai and Zhuang Liu and Qi Qian and Wei Wen and Hongyu Xu and Hanxian Huang and Ernie Chang and Yanbei Chen and Vikas Chandra and Raghuraman Krishnamoorthi and Mohamed Elhoseiny and Lemeng Wu},
  year={2026},
  eprint={2611.24161},
  archivePrefix={arXiv},
  url={https://arxiv.org/abs/2611.24161}
}