![[2602.20980] CrystaL: Spontaneous Emergence of Visual Latents in MLLMs](https://arxiv.org/static/browse/0.3.4/images/arxiv-logo-fb.png){kind=logo}
[2602.20980] CrystaL: Spontaneous Emergence of Visual Latents in MLLMs
Summary
The paper presents CrystaL, a novel framework for Multimodal Large Language Models (MLLMs) that enhances visual understanding by crystallizing latent representations without external supervision.
Why It Matters
As MLLMs continue to evolve, improving their ability to integrate visual and language data is crucial for applications in AI. CrystaL addresses limitations in existing models, potentially leading to advancements in fine-grained visual understanding and reasoning capabilities, which are essential for tasks like image recognition and natural language processing.
Key Takeaways
- CrystaL introduces a single-stage framework for processing images in MLLMs.
- It aligns attention patterns and predictions across intact and corrupted images.
- The framework enhances visual semantics without needing external annotations.
- Extensive experiments show CrystaL outperforms current state-of-the-art models.
- The approach maintains robust reasoning capabilities while improving visual understanding.
Computer Science > Computer Vision and Pattern Recognition arXiv:2602.20980 (cs) [Submitted on 24 Feb 2026] Title:CrystaL: Spontaneous Emergence of Visual Latents in MLLMs Authors:Yang Zhang, Danyang Li, Yuxuan Li, Xin Zhang, Tianyu Xie, Mingming Cheng, Xiang Li View a PDF of the paper titled CrystaL: Spontaneous Emergence of Visual Latents in MLLMs, by Yang Zhang and 6 other authors View PDF HTML (experimental) Abstract:Multimodal Large Language Models (MLLMs) have achieved remarkable performance by integrating powerful language backbones with large-scale visual encoders. Among these, latent Chain-of-Thought (CoT) methods enable implicit reasoning in continuous hidden states, facilitating seamless vision-language integration and faster inference. However, existing heuristically predefined supervision signals in latent CoT provide limited guidance for preserving critical visual information in intermediate latent states. To address this limitation, we propose CrystaL (Crystallized Latent Reasoning), a single-stage framework with two paths to process intact and corrupted images, respectively. By explicitly aligning the attention patterns and prediction distributions across the two paths, CrystaL crystallizes latent representations into task-relevant visual semantics, without relying on auxiliary annotations or external modules. Extensive experiments on perception-intensive benchmarks demonstrate that CrystaL consistently outperforms state-of-the-art baselines, achieving subs...
