dorsal/arxiv
View SchemaForest Before Trees: Latent Superposition for Efficient Visual Reasoning
| Authors | Yubo Wang, Juntian Zhang, Yichen Wu, Yankai Lin, Nils Lukas, Yuhan Liu |
|---|---|
| Categories | |
| ArXiv ID | 2601.06803vv1 |
| URL | https://arxiv.org/abs/2601.06803 |
| License | http://creativecommons.org/licenses/by-nc-nd/4.0/ |
Abstract
While Chain-of-Thought empowers Large Vision-Language Models with multi-step reasoning, explicit textual rationales suffer from an information bandwidth bottleneck, where continuous visual details are discarded during discrete tokenization. Recent latent reasoning methods attempt to address this challenge, but often fall prey to premature semantic collapse due to rigid autoregressive objectives. In this paper, we propose Laser, a novel paradigm that reformulates visual deduction via Dynamic Windowed Alignment Learning (DWAL). Instead of forcing a point-wise prediction, Laser aligns the latent state with a dynamic validity window of future semantics. This mechanism enforces a "Forest-before-Trees" cognitive hierarchy, enabling the model to maintain a probabilistic superposition of global features before narrowing down to local details. Crucially, Laser maintains interpretability via decodable trajectories while stabilizing unconstrained learning via Self-Refined Superposition. Extensive experiments on 6 benchmarks demonstrate that Laser achieves state-of-the-art performance among latent reasoning methods, surpassing the strong baseline Monet by 5.03% on average. Notably, it achieves these gains with extreme efficiency, reducing inference tokens by more than 97%, while demonstrating robust generalization to out-of-distribution domains.
{
"annotation_id": "62413b11-063f-4991-b86a-db221a945581",
"date_created": "2026-02-17T05:53:08.814000Z",
"date_modified": "2026-02-17T05:53:08.814000Z",
"file_hash": "9143697a065ed237c99d8c5f9474ac3605946ea48bb4542eb34bd6697d6eab8c",
"private": false,
"record": {
"abstract": "While Chain-of-Thought empowers Large Vision-Language Models with multi-step reasoning, explicit textual rationales suffer from an information bandwidth bottleneck, where continuous visual details are discarded during discrete tokenization. Recent latent reasoning methods attempt to address this challenge, but often fall prey to premature semantic collapse due to rigid autoregressive objectives. In this paper, we propose Laser, a novel paradigm that reformulates visual deduction via Dynamic Windowed Alignment Learning (DWAL). Instead of forcing a point-wise prediction, Laser aligns the latent state with a dynamic validity window of future semantics. This mechanism enforces a \"Forest-before-Trees\" cognitive hierarchy, enabling the model to maintain a probabilistic superposition of global features before narrowing down to local details. Crucially, Laser maintains interpretability via decodable trajectories while stabilizing unconstrained learning via Self-Refined Superposition. Extensive experiments on 6 benchmarks demonstrate that Laser achieves state-of-the-art performance among latent reasoning methods, surpassing the strong baseline Monet by 5.03% on average. Notably, it achieves these gains with extreme efficiency, reducing inference tokens by more than 97%, while demonstrating robust generalization to out-of-distribution domains.",
"arxiv_id": "2601.06803",
"authors": [
"Yubo Wang",
"Juntian Zhang",
"Yichen Wu",
"Yankai Lin",
"Nils Lukas",
"Yuhan Liu"
],
"categories": [
"cs.CL",
"cs.CV"
],
"license": "http://creativecommons.org/licenses/by-nc-nd/4.0/",
"title": "Forest Before Trees: Latent Superposition for Efficient Visual Reasoning",
"url": "https://arxiv.org/abs/2601.06803",
"version": "v1"
},
"schema_id": "dorsal/arxiv",
"source": {
"execution_id": "7947416e-d0a3-437f-b765-0b126f5b90a4",
"id": "arXiv Dataset",
"type": "Model",
"variant": "snapshot-2026-01-17",
"version": "0.1.0"
},
"user_id": 1000002
}