ResearchBenchmarksVisual Question Answering on ROBOSPATIALFor AI assistantsMarkdownContent indexVisual Question Answering on ROBOSPATIALFollow69.5AccuracyVALOR58.37261.26164.1567.039Dec 9, 2025Evaluation resultsMethodMethodLinksAccuracyVALORBase Model=Qwen3-8BBase Model=Qwen3-8B2025.1269.5ViGoRLBase Model=Qwen2.5-VL-7BBase Model=Qwen2.5-VL-7B2025.1264.9VALOR-RLBase Model=Qwen3-8BBase Model=Qwen3-8B2025.1261.8GRITBase Model=Qwen2.5-VL-3BBase Model=Qwen2.5-VL-3B2025.1258.8