Share your thoughts, 1 month free Claude Pro on usSee more
WorkDL logo mark

CFPO: Counterfactual Policy Optimization for Multimodal Reasoning

About

Large Vision-Language Models (LVLMs) have demonstrated remarkable capabilities in multimodal reasoning. However, prevailing reinforcement learning (RL) paradigms lack explicit counterfactual enhancement and causal learning mechanisms. This fundamental deficiency results in severe grounding failures, manifesting as a tendency to ignore visual evidence in favor of language priors or exhibiting hallucination drift during long chain-of-thought reasoning. To address this root cause, we propose CounterFactual Policy Optimization (CFPO), a novel framework that enforces causal consistency between visual perception and textual reasoning. CFPO introduces a cross-modal counterfactual enhancement mechanism, which regularizes the policy by maximizing the discrepancy between the model's predictions and those from a counterfactual state where critical visual cues are suppressed. This approach seamlessly integrates with standard algorithms like GRPO and DAPO without requiring external reward models or additional supervision. Extensive experiments demonstrate that CFPO significantly improves reasoning fidelity, achieving consistent gains of 3.17%-6.25% over standard RL baselines and 1.32%-2.13% over the state-of-the-art perception-aware method (PAPO). Code is available at https://github.com/Raven-July/CFPO.

Zhangyuan Yu, Wanran Sun, Guangjing Yang, Xiaohu Wu, Qicheng Lao• 2026

Related benchmarks

TaskDatasetResultRank
Mathematical ReasoningWeMath
Accuracy68.25
317
Mathematical ReasoningMathVerse
Accuracy61.92
266
Multimodal ReasoningWeMath
Accuracy68.79
199
Multimodal ReasoningLogicVista
Accuracy48.07
172
Logical reasoningLogicVista
Accuracy44.45
163
Multimodal ReasoningMathVerse
Accuracy68.47
138
Mathematical ReasoningMathVerse-V
Accuracy58.43
49
Mathematical ReasoningMathVista
Accuracy66.96
37
Multimodal ReasoningMMMU-Pro
Accuracy27.96
33
Multimodal ReasoningMMK12
Accuracy71.61
29
Showing 10 of 19 rows

Other info

Follow for update