dorsal/arxiv
View SchemaLaViT: Aligning Latent Visual Thoughts for Multi-modal Reasoning
| Authors | Linquan Wu, Tianxiang Jiang, Yifei Dong, Haoyu Yang, Fengji Zhang, Shichaang Meng, Ai Xuan, Linqi Song, Jacky Keung |
|---|---|
| Categories | |
| ArXiv ID | 2601.10129vv1 |
| URL | https://arxiv.org/abs/2601.10129 |
| License | http://creativecommons.org/licenses/by/4.0/ |
Abstract
Current multimodal latent reasoning often relies on external supervision (e.g., auxiliary images), ignoring intrinsic visual attention dynamics. In this work, we identify a critical Perception Gap in distillation: student models frequently mimic a teacher's textual output while attending to fundamentally divergent visual regions, effectively relying on language priors rather than grounded perception. To bridge this, we propose LaViT, a framework that aligns latent visual thoughts rather than static embeddings. LaViT compels the student to autoregressively reconstruct the teacher's visual semantics and attention trajectories prior to text generation, employing a curriculum sensory gating mechanism to prevent shortcut learning. Extensive experiments show that LaViT significantly enhances visual grounding, achieving up to +16.9% gains on complex reasoning tasks and enabling a compact 3B model to outperform larger open-source variants and proprietary models like GPT-4o.
{
"annotation_id": "fbb56814-879b-4531-9a70-f4918a93bd90",
"date_created": "2026-02-17T05:53:24.098000Z",
"date_modified": "2026-02-17T05:53:24.098000Z",
"file_hash": "50a531bda46c1c2cae519bc044b690126b829f8940a5f4a7d823af4c964b9d68",
"private": false,
"record": {
"abstract": "Current multimodal latent reasoning often relies on external supervision (e.g., auxiliary images), ignoring intrinsic visual attention dynamics. In this work, we identify a critical Perception Gap in distillation: student models frequently mimic a teacher\u0027s textual output while attending to fundamentally divergent visual regions, effectively relying on language priors rather than grounded perception. To bridge this, we propose LaViT, a framework that aligns latent visual thoughts rather than static embeddings. LaViT compels the student to autoregressively reconstruct the teacher\u0027s visual semantics and attention trajectories prior to text generation, employing a curriculum sensory gating mechanism to prevent shortcut learning. Extensive experiments show that LaViT significantly enhances visual grounding, achieving up to +16.9% gains on complex reasoning tasks and enabling a compact 3B model to outperform larger open-source variants and proprietary models like GPT-4o.",
"arxiv_id": "2601.10129",
"authors": [
"Linquan Wu",
"Tianxiang Jiang",
"Yifei Dong",
"Haoyu Yang",
"Fengji Zhang",
"Shichaang Meng",
"Ai Xuan",
"Linqi Song",
"Jacky Keung"
],
"categories": [
"cs.CV",
"cs.AI"
],
"license": "http://creativecommons.org/licenses/by/4.0/",
"title": "LaViT: Aligning Latent Visual Thoughts for Multi-modal Reasoning",
"url": "https://arxiv.org/abs/2601.10129",
"version": "v1"
},
"schema_id": "dorsal/arxiv",
"source": {
"execution_id": "1d931212-b0eb-42f3-a1be-05f4eba84692",
"id": "arXiv Dataset",
"type": "Model",
"variant": "snapshot-2026-01-17",
"version": "0.1.0"
},
"user_id": 1000002
}