AI Paper Atlas

Research map / Video / Video-language understanding

STVG-R1: Incentivizing Instance-Level Reasoning and Grounding in Videos via Reinforcement Learning

Xiaowen Zhang, Zhi Gao, Licheng Jiao, Lingling Li, Qing Li

ICLR 2026 · Video › Video-language understanding · approach: Temporal video grounding · 4 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
add object-centric visual prompts (zero-shot) (ablation)Qwen3-VL-8B · HCSTVG-v1 testvIoU@0.3: 28.2 -> 56.5Improved
add object-centric visual prompts (zero-shot) (ablation)Qwen2.5-VL-7B · HCSTVG-v1 testm_tIoU: 40.3 -> 38.7Hurt
add object-centric visual prompts (zero-shot) (ablation)Qwen2.5-VL-7B · HCSTVG-v1 testvIoU@0.3: 28.2 -> 40.7Improved
add object-centric visual prompts (zero-shot) (ablation)Qwen2.5-VL-7B · HCSTVG-v2 valm_vIoU: 19.3 -> 19.5No clear effect
add GRPO with temporal reward only (no visual prompts) (ablation)Qwen2.5-VL-7B · HCSTVG-v1 testm_tIoU: 40.3 -> 57.5Improved
add GRPO on top of visual prompts instead of SFT (ablation)Qwen2.5-VL-7B · HCSTVG-v1 testm_vIoU: 34.3 -> 39.1Improved
add visual-prompt SFT instead of GRPO (ablation)Qwen2.5-VL-7B · ST-Align STVGm_tIoU: 38.3 -> 36.6Hurt
use coupled spatio-temporal reward instead of sparse spatial reward (ablation)STVG-R1 · HCSTVG-v1 testm_vIoU: 39.1 -> 38.3Hurt
use continuous per-frame IoU spatial reward (ablation)STVG-R1 · HCSTVG-v1 testm_vIoU: 39.1 -> 38.6Hurt
use mixed number+letter prompts instead of numbers (ablation)zero-shot Qwen2.5-VL-7B · HCSTVG-v1m_vIoU: 24.9 -> 15.7Hurt
increase prompt font size from 20 to 40 (ablation)zero-shot Qwen2.5-VL-7B · HCSTVG-v1vIoU@0.3: 40.6 -> 36.4Hurt
raise mask filtering threshold to 1/2 from 1/3 (ablation)data quality and zero-shot · HCSTVG-v1m_vIoU (upper bound): 67.7 -> 65.1Hurt

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.

Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy