Progressive Reasoning with Primitive Correction for Compositional Zero-Shot Learning

2026-07-07

AI Summary

This work addresses the limitations in compositional zero-shot learning (CZSL) stemming from inadequate modeling of contextual dependencies between attributes and objects, as well as error propagation caused by unidirectional conditional reasoning. To overcome these issues, the authors propose the PRPC framework, which leverages a structured question-answering chain-of-thought to enable bidirectional conditional inference, allowing mutual correction during stepwise prediction to suppress early errors. The approach integrates multimodal large language models with a GRPO-based reinforcement learning post-training strategy, incorporating step-level rewards to enhance logical consistency. Evaluated on three CZSL benchmarks, the method achieves state-of-the-art performance, demonstrating the effectiveness of progressive bidirectional reasoning for compositional generalization.

Abstract

Compositional Zero-Shot Learning (CZSL) aims to combine known attributes and objects as primitives for recognizing previously unseen attribute-object pairs. Prior works either predict attributes and objects independently, missing their strong contextual dependency, or use unidirectional conditional modeling (e.g., object-guided attribute prediction), which is prone to error propagation. We propose PRPC, a Progressive Reasoning framework with Primitive Correction, which explicitly models the bidirectional dependency between attributes and objects via step-wise inference. PRPC performs mutual correction of primitives to suppress prediction errors in earlier steps. Specifically, we formulate CZSL as structured, Q&A-style Chain-of-Thought reasoning process and constrain the MLLM to follow predefined semantic steps to generate intermediate decisions. To further enhance the reliability and logical consistency of intermediate reasoning, we introduce reinforcement learning post-training with a GRPO-based objective, providing step-level rewards aligned with the progressive inference procedure. Extensive experiments on three CZSL benchmarks demonstrate that PRPC achieves state-of-the-art performance, validating the effectiveness of progressive reasoning and bidirectional correction for robust compositional generalization.

Problem

• Compositional Zero-Shot Learning
• attribute-object compositionality
• contextual dependency
• error propagation
• compositional generalization

Innovation

• Progressive Reasoning
• Primitive Correction
• Compositional Zero-Shot Learning
• Bidirectional Dependency
• Chain-of-Thought