AI Paper Atlas

Research map / Vision-language and multimodal / Robot action and navigation

Training-free Generation of Temporally Consistent Rewards from VLMs

Yinuo Zhao, Jiale Yuan, Zhiyuan Xu, Xiaoshuai Hao, Xinyi Zhang, Kun Wu, Zhengping Che, Chi Harold Liu, Jian Tang

ICCV 2025 · Vision-language and multimodal › Robot action and navigation · approach: Multimodal prompt-driven manipulation · 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 VLM initialization of particle filter instead of random initialization (ablation)T2-VLM · Place-same-color reward accuracyreward accuracy (%): 81.72 -> 94.60Improved
add particle filter instead of querying VLM at every step (ablation)T2-VLM · first three tasks reward accuracyreward accuracy (%): below 70%Improved
use T2-VLM instead of VLM-score with Claude-3.5-sonnetVLM-score · Place-same-colorreward accuracy (%): 54.77 -> 90.56Improved
use T2-VLM instead of VLM-score with Qwen2-VL-7BVLM-score · Place-same-colorreward accuracy (%): 42.08 -> 83.25Improved
use T2-VLM instead of VLM-score with Claude-3.5-sonnetVLM-score · Make-lineinference time (s): 166.20 -> 65.59Improved
use T2-VLM instead of VLM-score with Claude-3.5-sonnetVLM-score · Stack-towerVLM query times: 1620 -> 200Improved
use T2-VLM instead of VLM-score with Qwen2-VL-72BVLM-score · Make-lineinference time (s): 42.39 -> 42.10No clear effect
use larger VLM (Qwen2-VL-72B instead of Qwen2-VL-7B) in VLM-scoreVLM-score · Stack-towerreward accuracy (%): 76.96 -> 68.12Hurt
use RL with T2-VLM rewards instead of REFLECTREFLECT · recovery averagerecovery success rate: 0.53 -> 0.93Improved
use RL with T2-VLM rewards instead of SayCanSayCan · recovery averagerecovery meta steps: 18.53 -> 9.73Improved
reduce PF hidden length below 25 (ablation)T2-VLM · Place-same-colorreward accuracy (%)Hurt

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