Agent Explorative Policy Optimization for Multimodal Agentic Reasoning
Abstract
Agents using vision-language models with extended reasoning face challenges in tool utilization, which are addressed through AXPO, a method that improves performance by optimizing thinking prefixes and tool call resampling.
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.
Community
AXPO (Agent Explorative Policy Optimization) addresses the thinking-acting gap in multimodal agentic reasoning by resampling tool calls in failed rollouts to improve training signal and model performance.
Thank you for uploading our work, @taesiri !
We've also shared the key contributions in companion posts:
X: https://x.com/mkkang_1133/status/2059872464461848581?s=20
LinkedIn: https://www.linkedin.com/posts/minki-kang-5aa7281bb_releasing-axpo-our-new-work-on-agentic-activity-7465640179035279360-UZuh
Happy to discuss or take feedback :)
Models citing this paper 0
No model linking this paper
Datasets citing this paper 0
No dataset linking this paper
Spaces citing this paper 0
No Space linking this paper
