iReasoner: Trajectory-Aware Intrinsic Reasoning Supervision for Self-Evolving Large Multimodal Models

TL;DR

iReasoner introduces a novel self-evolving framework that enhances large multimodal models' reasoning capabilities by explicitly training intermediate reasoning steps using trajectory-aware intrinsic rewards, leading to improved performance without labeled data.

Contribution

The paper presents a new self-evolving approach that explicitly trains intermediate reasoning in LMMs using trajectory-aware signals, improving reasoning without external supervision.

Findings

Up to +2.1 points improvement on multimodal reasoning benchmarks.

Effective training of intermediate reasoning steps without ground-truth labels.

Code available at https://meghanaasunil.github.io/iReasoner.

Abstract

Recent work shows that large multimodal models (LMMs) can self-improve from unlabeled data via self-play and intrinsic feedback. Yet existing self-evolving frameworks mainly reward final outcomes, leaving intermediate reasoning weakly constrained despite its importance for visually grounded decision making. We propose iReasoner, a self-evolving framework that improves an LMM's implicit reasoning by explicitly eliciting chain-of-thought (CoT) and rewarding its internal agreement. In a Proposer--Solver loop over unlabeled images, iReasoner augments outcome-level intrinsic rewards with a trajectory-aware signal defined over intermediate reasoning steps, providing learning signals that distinguish reasoning paths leading to the same answer without ground-truth labels or external judges. Starting from Qwen2.5-VL-7B, iReasoner yields up to $+2.1$ points across diverse multimodal reasoning benchmarks under fully unsupervised post-training. We hope this work serves as a starting point for reasoning-aware self-improvement in LMMs in purely unsupervised settings. Our code is available at https://meghanaasunil.github.io/iReasoner.