AI Paper Atlas

Research map / Vision-language and multimodal / Multimodal LLMs

GeoSolver: Scaling Test-Time Reasoning in Remote Sensing with Fine-Grained Process Supervision

Sun Lang, Ronghao Fu, Zhuoran Duan, Haoran Liu, Xueyan Liu, Bo Yang

ECCV 2026 · Vision-language and multimodal › Multimodal LLMs · approach: Spatial reasoning in VLMs · 3 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 average PRM process score as reward (ablation)Vanilla GRPO · avg of 6 RS tasksaverage score: 67.09 -> 58.13Hurt
Add drop-moment (process-aware) penalty (ablation)Vanilla GRPO · avg of 6 RS tasksaverage score: 67.09 -> 68.16Improved
Add entropy-guided tree exploration (ablation)Vanilla GRPO · avg of 6 RS tasksaverage score: 67.09 -> 68.13Improved
Combine drop-moment penalty with tree exploration (ablation)Process-Aware Tree-GRPO · avg of 6 RS tasksaverage score: 67.09 -> 70.51Improved
Add SFT on Geo-CoT380k (ablation)GLM-4.1V-Base · avg of 6 RS tasksaverage score: 22.56 -> 60.34Improved
Remove MCTS-derived samples from GeoPRM training data (ablation)GeoPRM (GeoSolver, BoN N=32) · object-levelscore: 74.26 -> 64.27Hurt
Remove synthetic hallucination injection from GeoPRM training data (ablation)GeoPRM (GeoSolver, BoN N=32) · object-levelscore: 74.26 -> 70.26Hurt
Remove synthetic hallucination injection (ablation)GeoPRM (GeoSolver, BoN N=32) · scene-level N=4score: 74.03 -> 74.24No clear effect
Use lower drop threshold rho=0.2 instead of 0.3 (ablation)Drop-moment penalty · avg scoreaverage score: 70.51 -> 69.06Hurt
Use severe penalty gamma=0.3 instead of 0.7 (ablation)Drop-moment penalty · avg scoreaverage score: 70.51 -> 68.21Hurt
Use GeoPRM verification (BoN or beam search) vs greedy decodingGeoSolver · six RS tasksperformanceImproved
Use generic VLM or math PRM verifiers instead of GeoPRMverifier choice · RS tasksperformanceHurt

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 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