AI Paper Atlas

Research map / Vision-language and multimodal / Multimodal LLMs

Beyond Where to Look: Trajectory-Guided Reinforcement Learning for Multimodal RLVR

Jinda Lu, Junkang Wu, Jinghan Li, Kexin Huang, Shuo Yang, Mingzhu Chen, Jiancan Wu, Kuien Liu, Xiang Wang

ECCV 2026 · Vision-language and multimodal › Multimodal LLMs · approach: Reinforcement learning for visual reasoning · 1 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 trajectory filtering (keep only correct expert trajectories) (ablation)TGRL-GRPO · Avg of 5 benchmarks (Geo3K)average accuracy: 57.50 -> 58.46Improved
add token reweighting to expert trajectories (ablation)TGRL-GRPO · Avg of 5 benchmarks (Geo3K)average accuracy: 57.03 -> 58.46Improved
add expert trajectories (with reweighting and filtering) (ablation)GRPO · Avg of 5 benchmarks (Geo3K)average accuracy: 57.16 -> 58.46Improved
add expert trajectories without token reweighting (ablation)GRPO · Avg of 5 benchmarks (Geo3K)average accuracy: 57.16 -> 57.03Hurt
add trajectory filtering (ablation)TGRL-DAPO · Avg of 5 benchmarks (Geo3K)average accuracy: 57.93 -> 59.51Improved
add token reweighting (ablation)TGRL-DAPO · Avg of 5 benchmarks (Geo3K)average accuracy: 57.50 -> 59.51Improved
add expert trajectories (with reweighting and filtering) (ablation)DAPO · Avg of 5 benchmarks (Geo3K)average accuracy: 56.92 -> 59.51Improved
set token reweighting beta to 5 instead of 7 (ablation)TGRL-GRPO · Avg of 5 benchmarks (Geo3K)average accuracy: 57.36 -> 58.46Improved
set token reweighting beta to 14 instead of 15 (ablation)TGRL-DAPO · Avg of 5 benchmarks (Geo3K)average accuracy: 57.88 -> 59.51Improved
add TGRL trajectories to GRPO with 39K ViRL39K training dataGRPO · Avg of 5 benchmarks (ViRL39K)average accuracy: 57.71 -> 59.38Improved
add TGRL trajectories to DAPO with 39K ViRL39K training dataDAPO · Avg of 5 benchmarks (ViRL39K)average accuracy: 58.15 -> 60.68Improved
scale TGRL-GRPO training data from 2.1K Geo3K to 39K ViRL39KTGRL-GRPO · Avg of 5 benchmarksaverage accuracy: 58.46 -> 59.38Improved

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