Hugging Face Daily PapersNishad Singhi, Hector Garcia Rodriguez, Aditya AroraSep 14, 20261 min readpaperadvanced

Reasoning with Image Generation

Summary

ReImaGin proposes using image generation models as a flexible visual reasoning mechanism for multimodal LLMs, moving beyond rigid visual tools. It consistently outperforms text-only and specialist vision-tool baselines by up to 25% on diverse visual reasoning tasks.

Chain-of-thought reasoning in LLMs is limited when tasks require direct visual manipulation.
Existing multimodal LLMs use narrow, rigid visual expert tools that cannot generate or transform content.
ReImaGin leverages image generation models for flexible, open-ended visual operations.
It enables complex visual tasks like removing occlusions or generating floorplans from multiple views.

This paper is important for AI researchers and engineers developing multimodal LLMs, as it offers a new paradigm for visual reasoning that overcomes limitations of current fixed-function tools.

8/10

llm machine learning computer vision reasoning image generation multimodal ai

Related reading

OmniHarness: Harnessing Generalizable Visual Generation via Symbolic Policy Learning arxiv.org
OmniHarness introduces a symbolic‑policy framework that extracts reusable procedural knowledge from multimodal LLM‑driven visual generation runs. By decoupling task logic from instance inputs, the system can instantiate, adapt, and compose policies for new visual tasks, using intermediate verification for on‑the‑fly refinement while keeping the underlying model frozen. Self‑directed practice task…
Hugging Face Daily Papersarxiv.orgXu Xu, Jinxiu Liu, Zhangbo Qiao23d1 minpaper
Scaffolding Minds: Optimizing Latent Visual Target Representations for Multimodal Reasoning arxiv.org
The paper pinpoints two weaknesses in latent‑reasoning pipelines—suboptimal visual token encoders and deterministic RL sampling—and proposes Scaffolding Minds, which trains a dedicated scaffolding encoder and a stochastic RL sampler (learning both mean and variance). This yields up to +19 points improvement on spatial planning tasks and +5.6 points average gain on nine visual reasoning benchmarks.
Hugging Face Daily Papersarxiv.orgHaoqiang Kang, Yinpeng Chen, Luyang Liu7d1 minpaper
Imagine3D-LLM: Teaching MLLMs to Imagine 3D Scenes Before Answering arxiv.org
Imagine3D-LLM augments a multimodal LLM with a small set of learnable summary tokens that are decoded into a compact 3D Gaussian splatting of the scene, supervised by photometric reconstruction. This auxiliary task improves cross‑view reasoning and yields better results on spatial‑reasoning benchmarks.
Hugging Face Daily Papersarxiv.orgJaewoo Jung, Hyeonseo Yu, Honggyu An7d1 minpaper
ThinkV2V: Unleashing the Reasoning Capability of MLLMs for Instruction-Guided Video Editing arxiv.org
ThinkV2V adds an explicit reasoning step using a multimodal LLM before conditioning a video diffusion model, trained with a progressive curriculum and refined at inference time. The 5B DiT model trained on a new 150K dataset outperforms larger baselines on complex instruction-guided edits.
Hugging Face Daily Papersarxiv.orgDonghao Zhou, Haoyang He, Fan Zhang7d1 minpaper
Can MiniMax-H3 Reason About the Physical World? An Evaluation of Omni-Modal Generative Model arxiv.org
This paper introduces a novel evaluation framework to assess the physical world reasoning capabilities of omni-modal generative models like MiniMax-H3. It found that MiniMax-H3 achieved an overall success rate of 41.97% across 517 instances, with significant performance variations depending on the input modalities and reasoning tasks.
Hugging Face Daily Papersarxiv.orgHaoyu Zhao, Zihao Zhao, Tianyu Deng20d2 minpaper
AutoRef: Harness Optimization for Agentic Multi-Reference Image Generation arxiv.org
AutoRef uses a coding LLM to automatically rewrite the harness that orchestrates multi‑reference image generation, keeping the generator frozen. The resulting harness lifts FLUX.2‑4B scores from 5.72 to 7.37 on MultiBanana and generalizes across models and benchmarks.
Hugging Face Daily Papersarxiv.orgYuta Oshima, Ku Onoda, Yusuke Iwasawa8d2 minpaper

Reasoning with Image Generation

Summary

Related reading

OmniHarness: Harnessing Generalizable Visual Generation via Symbolic Policy Learningarxiv.org

Scaffolding Minds: Optimizing Latent Visual Target Representations for Multimodal Reasoningarxiv.org

Imagine3D-LLM: Teaching MLLMs to Imagine 3D Scenes Before Answeringarxiv.org

ThinkV2V: Unleashing the Reasoning Capability of MLLMs for Instruction-Guided Video Editingarxiv.org

Can MiniMax-H3 Reason About the Physical World? An Evaluation of Omni-Modal Generative Modelarxiv.org

AutoRef: Harness Optimization for Agentic Multi-Reference Image Generationarxiv.org

OmniHarness: Harnessing Generalizable Visual Generation via Symbolic Policy Learning arxiv.org

Scaffolding Minds: Optimizing Latent Visual Target Representations for Multimodal Reasoning arxiv.org

Imagine3D-LLM: Teaching MLLMs to Imagine 3D Scenes Before Answering arxiv.org

ThinkV2V: Unleashing the Reasoning Capability of MLLMs for Instruction-Guided Video Editing arxiv.org

Can MiniMax-H3 Reason About the Physical World? An Evaluation of Omni-Modal Generative Model arxiv.org

AutoRef: Harness Optimization for Agentic Multi-Reference Image Generation arxiv.org