Saved in:
Bibliographic Details
Main Authors: Zhang, Xiaowen, Gao, Zhi, Jiao, Licheng, Li, Lingling, Li, Qing
Format: Preprint
Published: 2026
Subjects:
Online Access:https://arxiv.org/abs/2602.11730
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866910020108550144
author Zhang, Xiaowen
Gao, Zhi
Jiao, Licheng
Li, Lingling
Li, Qing
author_facet Zhang, Xiaowen
Gao, Zhi
Jiao, Licheng
Li, Lingling
Li, Qing
contents In vision-language models (VLMs), misalignment between textual descriptions and visual coordinates often induces hallucinations. This issue becomes particularly severe in dense prediction tasks such as spatial-temporal video grounding (STVG). Prior approaches typically focus on enhancing visual-textual alignment or attaching auxiliary decoders. However, these strategies inevitably introduce additional trainable modules, leading to significant annotation costs and computational overhead. In this work, we propose a novel visual prompting paradigm that avoids the difficult problem of aligning coordinates across modalities. Specifically, we reformulate per-frame coordinate prediction as a compact instance-level identification problem by assigning each object a unique, temporally consistent ID. These IDs are embedded into the video as visual prompts, providing explicit and interpretable inputs to the VLMs. Furthermore, we introduce STVG-R1, the first reinforcement learning framework for STVG, which employs a task-driven reward to jointly optimize temporal accuracy, spatial consistency, and structural format regularization. Extensive experiments on six benchmarks demonstrate the effectiveness of our approach. STVG-R1 surpasses the baseline Qwen2.5-VL-7B by a remarkable margin of 20.9% on m_IoU on the HCSTVG-v2 benchmark, establishing a new state of the art (SOTA). Surprisingly, STVG-R1 also exhibits strong zero-shot generalization to multi-object referring video object segmentation tasks, achieving a SOTA 47.3% J&F on MeViS.
format Preprint
id arxiv_https___arxiv_org_abs_2602_11730
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle STVG-R1: Incentivizing Instance-Level Reasoning and Grounding in Videos via Reinforcement Learning
Zhang, Xiaowen
Gao, Zhi
Jiao, Licheng
Li, Lingling
Li, Qing
Computer Vision and Pattern Recognition
In vision-language models (VLMs), misalignment between textual descriptions and visual coordinates often induces hallucinations. This issue becomes particularly severe in dense prediction tasks such as spatial-temporal video grounding (STVG). Prior approaches typically focus on enhancing visual-textual alignment or attaching auxiliary decoders. However, these strategies inevitably introduce additional trainable modules, leading to significant annotation costs and computational overhead. In this work, we propose a novel visual prompting paradigm that avoids the difficult problem of aligning coordinates across modalities. Specifically, we reformulate per-frame coordinate prediction as a compact instance-level identification problem by assigning each object a unique, temporally consistent ID. These IDs are embedded into the video as visual prompts, providing explicit and interpretable inputs to the VLMs. Furthermore, we introduce STVG-R1, the first reinforcement learning framework for STVG, which employs a task-driven reward to jointly optimize temporal accuracy, spatial consistency, and structural format regularization. Extensive experiments on six benchmarks demonstrate the effectiveness of our approach. STVG-R1 surpasses the baseline Qwen2.5-VL-7B by a remarkable margin of 20.9% on m_IoU on the HCSTVG-v2 benchmark, establishing a new state of the art (SOTA). Surprisingly, STVG-R1 also exhibits strong zero-shot generalization to multi-object referring video object segmentation tasks, achieving a SOTA 47.3% J&F on MeViS.
title STVG-R1: Incentivizing Instance-Level Reasoning and Grounding in Videos via Reinforcement Learning
topic Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2602.11730