iReasoner: Trajectory-Aware Intrinsic Reasoning Supervision for Self-Evolving Large Multimodal Models
Fuente:
arXiv
Enregistré dans:
| Auteurs principaux: | , , |
|---|---|
| Format: | Preprint |
| Publié: |
2026
|
| Sujets: | |
| Accès en ligne: | |
| Tags: |
Ajouter un tag
Pas de tags, Soyez le premier à ajouter un tag!
|
| _version_ | 1866916029726195712 |
|---|---|
| author | Sunil, Meghana Venmathimaran, Manikandarajan Kavitha, Muthu Subash |
| author_facet | Sunil, Meghana Venmathimaran, Manikandarajan Kavitha, Muthu Subash |
| contents | 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. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2601_05877 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | iReasoner: Trajectory-Aware Intrinsic Reasoning Supervision for Self-Evolving Large Multimodal Models Sunil, Meghana Venmathimaran, Manikandarajan Kavitha, Muthu Subash Computation and Language 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. |
| title | iReasoner: Trajectory-Aware Intrinsic Reasoning Supervision for Self-Evolving Large Multimodal Models |
| topic | Computation and Language |
| url | https://arxiv.org/abs/2601.05877 |