AIMay 26, 2025

Unveiling the Compositional Ability Gap in Vision-Language Reasoning Model

arXiv:2505.19406v18 citationsh-index: 4
Originality Incremental advance
AI Analysis

This work addresses the problem of compositional reasoning gaps in VLMs for AI researchers, providing incremental insights into training strategies.

The study investigated whether vision-language models (VLMs) can inherit compositional reasoning capabilities through post-training strategies like reinforcement learning (RL), finding that RL-trained models outperform supervised fine-tuning on compositional generalization but still struggle with cross-modal and cross-task scenarios, with improvements seen from techniques like caption-before-thinking and visual grounding.

While large language models (LLMs) demonstrate strong reasoning capabilities utilizing reinforcement learning (RL) with verifiable reward, whether large vision-language models (VLMs) can directly inherit such capabilities through similar post-training strategies remains underexplored. In this work, we conduct a systematic compositional probing study to evaluate whether current VLMs trained with RL or other post-training strategies can compose capabilities across modalities or tasks under out-of-distribution conditions. We design a suite of diagnostic tasks that train models on unimodal tasks or isolated reasoning skills, and evaluate them on multimodal, compositional variants requiring skill integration. Through comparisons between supervised fine-tuning (SFT) and RL-trained models, we identify three key findings: (1) RL-trained models consistently outperform SFT on compositional generalization, demonstrating better integration of learned skills; (2) although VLMs achieve strong performance on individual tasks, they struggle to generalize compositionally under cross-modal and cross-task scenario, revealing a significant gap in current training strategies; (3) enforcing models to explicitly describe visual content before reasoning (e.g., caption-before-thinking), along with rewarding progressive vision-to-text grounding, yields notable gains. It highlights two essential ingredients for improving compositionality in VLMs: visual-to-text alignment and accurate visual grounding. Our findings shed light on the current limitations of RL-based reasoning VLM training and provide actionable insights toward building models that reason compositionally across modalities and tasks.

Code Implementations1 repo
Foundations

The foundational work for this paper's niche, ranked by how specifically the neighbourhood builds on it — not by global fame.

Your Notes