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

Agent Explorative Policy Optimization for Multimodal Agentic Reasoning

About

Vision-language models with extended reasoning succeed on complex problems, but many real-world problems require external tools that internal reasoning alone often cannot resolve. Agentic reasoning therefore interleaves two behaviors with a structural asymmetry: thinking (the self-contained default) and tool use (a high-variance auxiliary acting). We refer to this asymmetry as the Thinking-Acting Gap. Under standard RL recipes like GRPO, the gap manifests as two diagnostic symptoms during training: tool use is attempted on only ~30% of rollouts, and when attempted, the tool-using rollouts within a group are all-wrong on ~40% of questions, suppressing the learning signal at the tool calls that needed it. We propose AXPO (Agent eXplorative Policy Optimization): for each all-wrong tool-using subgroup, AXPO fixes the thinking prefix and resamples the tool call and its continuation, paired with uncertainty-based prefix selection. Across nine multimodal benchmarks and three scales of Qwen3-VL-Thinking, SFT+AXPO outperforms SFT+GRPO at average (+1.8pp Pass@1 and +1.8pp Pass@4 at 8B on average) and 8B with SFT+AXPO surpasses the 32B Base on Pass@4 with 4 times fewer parameters.

Minki Kang, Shizhe Diao, Ryo Hachiuma, Sung Ju Hwang, Pavlo Molchanov, Yu-Chiang Frank Wang, Byung-Kwan Lee• 2026

Related benchmarks

TaskDatasetResultRank
High-resolution Visual UnderstandingHR-Bench-8K--
83
ReasoningMATH-Vision
Pass@160.6
32
Fine-Grained PerceptionHRBench 4K
Pass@183.3
26
Mathematical ReasoningDynaMath
Pass@179
25
PerceptionVisual Probe
Pass@145.8
16
PerceptionVisual Probe
Pass@467.9
16
PerceptionHRBen 4K
Pass@491
16
PerceptionHRBen 8K
Pass@490
16
ReasoningMATH-Vision
Pass@475.7
16
SearchHR-MM Search
Pass@125.9
16
Showing 10 of 22 rows

Other info

Follow for update