# See2Think: Do Multimodal Models Really Use Intermediate Visual States?

arXiv: https://arxiv.org/abs/2607.26769 · HTML: https://mlsanity.com/paper/2607.26769/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Placed in Multimodal reasoning, one of 17 directions in LLM reasoning.
- Builds on 19 of its 32 references that are in the map.
- First posted on arXiv 29 Jul 2026.

## Placed in

- LLM reasoning / [Multimodal reasoning](https://mlsanity.com/lr/multimodal-reasoning/)
- Multimodal LLMs / vision-language models / [VLM benchmarks and visual reasoning](https://mlsanity.com/mllm/vlm-benchmarks-and-visual-reasoning/)

## Builds on (most cited first, 19 of 19)

- [Multimodal Chain-of-Thought Reasoning in Language Models](https://mlsanity.com/paper/2302.00923/) (2023)
- [DeepEyes: Incentivizing "Thinking with Images" via Reinforcement Learning](https://mlsanity.com/paper/2505.14362/) (2025)
- [Visual Sketchpad: Sketching as a Visual Chain of Thought for Multimodal Language Models](https://mlsanity.com/paper/2406.09403/) (2024)
- [OpenThinkIMG: Learning to Think with Images via Visual Tool Reinforcement Learning](https://mlsanity.com/paper/2505.08617/) (2025)
- [MathCanvas: Intrinsic Visual Chain-of-Thought for Multimodal Mathematical Reasoning](https://mlsanity.com/paper/2510.14958/) (2025)
- [V-Thinker: Interactive Thinking with Images](https://mlsanity.com/paper/2511.04460/) (2025)
- [When Visualizing is the First Step to Reasoning: MIRA, a Benchmark for Visual Chain-of-Thought](https://mlsanity.com/paper/2511.02779/) (2025)
- [On the Faithfulness of Visual Thinking: Measurement and Enhancement](https://mlsanity.com/paper/2510.23482/) (2025)
- [ViC-Bench: Benchmarking Visual-Interleaved Chain-of-Thought Capability in MLLMs with Free-Style Intermediate State Representations](https://mlsanity.com/paper/2505.14404/) (2025)
- [What, Whether and How? Unveiling Process Reward Models for Thinking with Images Reasoning](https://mlsanity.com/paper/2602.08346/) (2026)
- [Reliable Thinking with Images](https://mlsanity.com/paper/2602.12916/) (2026)
- [Walk the Talk: Bridging the Reasoning-Action Gap for Thinking with Images via Multimodal Agentic Policy Optimization](https://mlsanity.com/paper/2604.06777/) (2026)
- [S1-VL: Scientific Multimodal Reasoning Model with Thinking-with-Images](https://mlsanity.com/paper/2604.21409/) (2026)
- [Thinking with Images as Continuous Actions: Numerical Visual Chain-of-Thought](https://mlsanity.com/paper/2602.23959/) (2026)
- [Let's Think with Images Efficiently! An Interleaved-Modal Chain-of-Thought Reasoning Framework with Dynamic and Precise Visual Thoughts](https://mlsanity.com/paper/2603.21754/) (2026)
- [Do Multimodal Agents Really Benefit from Tool Use? A Systematic Study of Capability Gains](https://mlsanity.com/paper/2606.02357/) (2026)
- [TVI-CoT: Text-Visual Interleaved Chain-of-Thought Reasoning for Multimodal Understanding](https://mlsanity.com/paper/2606.08464/) (2026)
- [TwiFF (Think With Future Frames): A Large-Scale Dataset for Dynamic Visual Reasoning](https://mlsanity.com/paper/2602.10675/) (2026)
- [How and What to Imagine? Visual Thinking in Unified Multimodal Models for Cross-View Spatial Reasoning](https://mlsanity.com/paper/2605.27310/) (2026)

## Abstract (arXiv)

Multimodal large language models increasingly use sketches, annotations, tools, and intermediate images during reasoning, but it remains unclear whether they truly rely on these visual states. Existing benchmarks are limited both by task collections with narrow coverage or partially text-solvable samples and by evaluations that emphasize final answers without diagnosing how intermediate visual states are generated, rendered, and used. We introduce See2Think, a unified evaluation framework comprising See2ThinkBench and Visual Action-of-Thought (VAoT). See2ThinkBench contains 1,200 open-ended, visually dependent problems across 12 task categories spanning 2D structured, 3D scene, and real-world reasoning. VAoT records textual thoughts, visual actions, rendered states, and subsequent reasoning under four controlled inference settings. Evaluating representative proprietary and open-source multimodal models, we find that visual reasoning is strongly model- and environment-dependent, with no single setting consistently dominating across tasks. Process analysis further shows that models usually select relevant visual operations, while faithful rendering remains the clearest bottleneck and high feedback uptake does not necessarily translate into accuracy gains. Under task-relevant corrupted feedback, models exhibit behavioral dependence on visual states, with accuracy dropping by over 10 percentage points in controlled interventions.