ResearchBenchmarksVisual Question Answering on TextVQA (F1, Hallucination Rate, Budget Usage)For AI assistantsMarkdownContent indexVisual Question Answering on TextVQA (F1, Hallucination Rate, Budget Usage)Follow72.9F1 ScorePoP64.37266.58668.871.014Feb 27, 2026Evaluation resultsMethodMethodLinksF1 ScoreHallucination RateAverage Budget Usage (B-bar)PoPReasoning protocol=PoPReasoning protocol=PoP2026.0272.99.312.4ProgVLMReasoning protocol=Pro...Reasoning protocol=ProgVLM2026.0269.715.114.7MM-ReActReasoning protocol=MM-...Reasoning protocol=MM-ReAct2026.0269.215.615.6M-CoTReasoning protocol=M-CoTReasoning protocol=M-CoT2026.0266.919.80.25DirectReasoning protocol=DirectReasoning protocol=Direct2026.0264.722.30.25