Beyond Visual CoT: Internalized Visual Thinking for Proactive Video Reasoning

Hugging Face Daily Papers Papers
multimodal-learning visual-reasoning video-ai inference-efficiency chain-of-thought post-training

Summary

The paper introduces Internalized Visual Thinking (IVT), a post-training framework that trains multimodal models to predict future frame embeddings, enabling direct answer generation without synthesizing intermediate images, thereby reducing latency over 5x for proactive video reasoning.

Multimodal large language models increasingly use visual chain-of-thought (Visual CoT) to reason about spatial, temporal, and embodied environments. By generating intermediate reasoning images, Visual CoT provides an intuitive mechanism for visual foresight but introduces substantial inference overhead, which is particularly problematic for proactive video reasoning. We ask whether models can learn to think visually during training while reasoning directly at inference. We introduce Internalized Visual Thinking (IVT), a post-training framework that jointly optimizes textual prediction and next-embedding prediction over unlabeled videos. Given a partially observed video, IVT predicts latent representations of future frames together with the target textual answer, encouraging the model to capture motion, object transitions, interactions, and latent intent. At inference, IVT generates the answer directly without synthesizing or re-encoding future frames. We conduct controlled studies across target representations, decoder designs, prediction horizons, data mixtures, training curricula, and predictive objectives. IVT improves over direct-answer fine-tuning on all six evaluation settings while retaining the same inference pathway. Compared with explicit Visual CoT, IVT achieves comparable or better performance and reduces average end-to-end latency by more than 5x. Together, our findings suggest that explicit pixel-space generation at inference time, as used in visual chain-of-thought, may not be necessary for effective proactive video reasoning. Predictive world modeling can be internalized during training to produce multimodal reasoners that are both more accurate and substantially more efficient.
Original Article
View Cached Full Text

Cached at: 08/18/26, 07:53 PM

Paper page - Beyond Visual CoT: Internalized Visual Thinking for Proactive Video Reasoning

Source: https://huggingface.co/papers/2608.15869

Abstract

Internalized Visual Thinking trains multimodal models to predict future frame embeddings during post-training, enabling direct answer generation at inference without synthesizing intermediate images and cutting latency over fivefold.

Multimodal large language modelsincreasingly use visual chain-of-thought (Visual CoT) to reason about spatial, temporal, and embodied environments. By generating intermediate reasoning images,Visual CoTprovides an intuitive mechanism for visual foresight but introduces substantial inference overhead, which is particularly problematic forproactive video reasoning. We ask whether models can learn to think visually during training while reasoning directly at inference. We introduceInternalized Visual Thinking(IVT), a post-training framework that jointly optimizes textual prediction andnext-embedding predictionover unlabeled videos. Given a partially observed video, IVT predicts latent representations of future frames together with the target textual answer, encouraging the model to capture motion, object transitions, interactions, and latent intent. At inference, IVT generates the answer directly without synthesizing or re-encoding future frames. We conduct controlled studies across target representations, decoder designs, prediction horizons, data mixtures, training curricula, and predictive objectives. IVT improves over direct-answer fine-tuning on all six evaluation settings while retaining the same inference pathway. Compared with explicitVisual CoT, IVT achieves comparable or better performance and reduces average end-to-end latency by more than 5x. Together, our findings suggest that explicit pixel-space generation at inference time, as used in visual chain-of-thought, may not be necessary for effectiveproactive video reasoning.Predictive world modelingcan be internalized during training to produce multimodal reasoners that are both more accurate and substantially more efficient.

View arXiv pageView PDFProject pageAdd to collection

Get this paper in your agent:

hf papers read 2608\.15869

Don’t have the latest CLI?curl \-LsSf https://hf\.co/cli/install\.sh \| bash

Models citing this paper0

No model linking this paper

Cite arxiv.org/abs/2608.15869 in a model README.md to link it from this page.

Datasets citing this paper0

No dataset linking this paper

Cite arxiv.org/abs/2608.15869 in a dataset README.md to link it from this page.

Spaces citing this paper0

No Space linking this paper

Cite arxiv.org/abs/2608.15869 in a Space README.md to link it from this page.

Collections including this paper0

No Collection including this paper

Add this paper to acollectionto link it from this page.

Similar Articles

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

Hugging Face Daily Papers

ThinkV2V introduces a reasoning-driven framework that activates MLLM thinking before visual generation for instruction-guided video editing, using an MLLM-to-DiT architecture with progressive curriculum training and inference-time thinking scaling. The authors also release the ThinkV2V-150K dataset and ThinkV2V-Bench, showing their 5B DiT model outperforms larger 10B baselines.