Semantic-Enriched Latent Visual Reasoning
Fuente:
arXiv
Gespeichert in:
| Hauptverfasser: | , , , , , , , , , , |
|---|---|
| Format: | Preprint |
| Veröffentlicht: |
2026
|
| Schlagworte: | |
| Online-Zugang: | |
| Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
| _version_ | 1866911722490560512 |
|---|---|
| author | Xu, Tianrun Sun, Yue Wang, Qixun Lu, Jingyi Wang, Yuan Zhang, Tianren Guo, Longteng Rao, Fengyun Lyu, Jing Chen, Feng Liu, Jing |
| author_facet | Xu, Tianrun Sun, Yue Wang, Qixun Lu, Jingyi Wang, Yuan Zhang, Tianren Guo, Longteng Rao, Fengyun Lyu, Jing Chen, Feng Liu, Jing |
| contents | Multimodal latent-space reasoning aims to replace explicit thinking with images by performing visual reasoning directly in a compact latent space. However, existing approaches largely rely on visual supervision and produce latent representations that lack sufficient semantic richness, limiting their ability to support diverse region-level reasoning tasks. In this work, we introduce Semantic-Enriched Latent Visual Reasoning (SLVR), a two-stage learning framework that enriches latent representations with attribute-level visual semantics and aligns them with diverse reasoning objectives. In the first stage, SLVR learns semantically enriched region-centric latents under fine-grained attribute supervision. In the second stage, we design Multi-query Group Relative Policy Optimization (M-GRPO) to align latent representations across multiple queries grounded in the same region. To support this framework, we construct SLV-Set, comprising approximately 400K region-level attribute annotations and 800K multi-query question answering samples, and introduce SV-QA, a benchmark that evaluates latent reasoning under semantic variation. Experiments demonstrate that SLVR improves the robustness and semantic consistency of latent visual reasoning compared to existing baselines. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2605_19342 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | Semantic-Enriched Latent Visual Reasoning Xu, Tianrun Sun, Yue Wang, Qixun Lu, Jingyi Wang, Yuan Zhang, Tianren Guo, Longteng Rao, Fengyun Lyu, Jing Chen, Feng Liu, Jing Computer Vision and Pattern Recognition Multimodal latent-space reasoning aims to replace explicit thinking with images by performing visual reasoning directly in a compact latent space. However, existing approaches largely rely on visual supervision and produce latent representations that lack sufficient semantic richness, limiting their ability to support diverse region-level reasoning tasks. In this work, we introduce Semantic-Enriched Latent Visual Reasoning (SLVR), a two-stage learning framework that enriches latent representations with attribute-level visual semantics and aligns them with diverse reasoning objectives. In the first stage, SLVR learns semantically enriched region-centric latents under fine-grained attribute supervision. In the second stage, we design Multi-query Group Relative Policy Optimization (M-GRPO) to align latent representations across multiple queries grounded in the same region. To support this framework, we construct SLV-Set, comprising approximately 400K region-level attribute annotations and 800K multi-query question answering samples, and introduce SV-QA, a benchmark that evaluates latent reasoning under semantic variation. Experiments demonstrate that SLVR improves the robustness and semantic consistency of latent visual reasoning compared to existing baselines. |
| title | Semantic-Enriched Latent Visual Reasoning |
| topic | Computer Vision and Pattern Recognition |
| url | https://arxiv.org/abs/2605.19342 |