# OmniOPSD: Rationale-Privileged On-Policy Self-Distillation for Affective Computing

arXiv: https://arxiv.org/abs/2606.15920 · HTML: https://mlsanity.com/paper/2606.15920/

## Key facts

- Cited by 6 papers in the map (data through 14 Oct 2026).
- Placed in Multimodal emotion understanding, one of 16 directions in Multimodal LLMs / vision-language models.
- Builds on 46 of its 57 references that are in the map.
- First posted on arXiv 14 Jun 2026.

## Placed in

- Multimodal LLMs / vision-language models / [Multimodal emotion understanding](https://mlsanity.com/mllm/multimodal-emotion-understanding/)
- Knowledge distillation / [Multimodal distillation](https://mlsanity.com/distill/multimodal-distillation/)

## Builds on (most cited first, 30 of 46)

- [Training language models to follow instructions with human feedback](https://mlsanity.com/paper/2203.02155/) (2022)
- [DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Models](https://mlsanity.com/paper/2402.03300/) (2024)
- [Direct Preference Optimization: Your Language Model is Secretly a Reward Model](https://mlsanity.com/paper/2305.18290/) (2023)
- [Visual Instruction Tuning](https://mlsanity.com/paper/2304.08485/) (2023)
- [Qwen2-VL: Enhancing Vision-Language Model's Perception of the World at Any Resolution](https://mlsanity.com/paper/2409.12191/) (2024)
- [Deep reinforcement learning from human preferences](https://mlsanity.com/paper/1706.03741/) (2017)
- [On-Policy Distillation of Language Models: Learning from Self-Generated Mistakes](https://mlsanity.com/paper/2306.13649/) (2023)
- [MELD: A Multimodal Multi-Party Dataset for Emotion Recognition in Conversations](https://mlsanity.com/paper/1810.02508/) (2018)
- [Self-Distilled Reasoner: On-Policy Self-Distillation for Large Language Models](https://mlsanity.com/paper/2601.18734/) (2026)
- [Reinforcement Learning via Self-Distillation](https://mlsanity.com/paper/2601.20802/) (2026)
- [Large Language Models Are Reasoning Teachers](https://mlsanity.com/paper/2212.10071/) (2022)
- [R1-VL: Learning to Reason with Multimodal Large Language Models via Step-wise Group Relative Policy Optimization](https://mlsanity.com/paper/2503.12937/) (2025)
- [SFT or RL? An Early Investigation into Training R1-Like Reasoning Large Vision-Language Models](https://mlsanity.com/paper/2504.11468/) (2025)
- [MOSI: Multimodal Corpus of Sentiment Intensity and Subjectivity Analysis in Online Opinion Videos](https://mlsanity.com/paper/1606.06259/) (2016)
- [A Survey of On-Policy Distillation for Large Language Models](https://mlsanity.com/paper/2604.00626/) (2026)
- [DFEW: A Large-Scale Database for Recognizing Dynamic Facial Expressions in the Wild](https://mlsanity.com/paper/2008.05924/) (2020)
- [Emotion-LLaMA: Multimodal Emotion Recognition and Reasoning with Instruction Tuning](https://mlsanity.com/paper/2406.11161/) (2024)
- [Insight-V: Exploring Long-Chain Visual Reasoning with Multimodal Large Language Models](https://mlsanity.com/paper/2411.14432/) (2024)
- [Why Does Self-Distillation (Sometimes) Degrade the Reasoning Capability of LLMs?](https://mlsanity.com/paper/2603.24472/) (2026)
- [R1-Omni: Explainable Omni-Multimodal Emotion Recognition with Reinforcement Learning](https://mlsanity.com/paper/2503.05379/) (2025)
- [Vision-OPD: Learning to See Fine Details for Multimodal LLMs via On-Policy Self-Distillation](https://mlsanity.com/paper/2605.18740/) (2026)
- [MAFW: A Large-scale, Multi-modal, Compound Affective Database for Dynamic Facial Expression Recognition in the Wild](https://mlsanity.com/paper/2208.00847/) (2022)
- [T-SciQ: Teaching Multimodal Chain-of-Thought Reasoning via Mixed Large Language Model Signals for Science Question Answering](https://mlsanity.com/paper/2305.03453/) (2023)
- [Make Acoustic and Visual Cues Matter: CH-SIMS v2.0 Dataset and AV-Mixup Consistent Module](https://mlsanity.com/paper/2209.02604/) (2022)
- [Towards the Law of Capacity Gap in Distilling Language Models](https://mlsanity.com/paper/2311.07052/) (2023)
- [AffectGPT: A New Dataset, Model, and Benchmark for Emotion Understanding with Multimodal Large Language Models](https://mlsanity.com/paper/2501.16566/) (2025)
- [EmoVIT: Revolutionizing Emotion Insights with Visual Instruction Tuning](https://mlsanity.com/paper/2404.16670/) (2024)
- [MIntRec: A New Dataset for Multimodal Intent Recognition](https://mlsanity.com/paper/2209.04355/) (2022)
- [From Static to Dynamic: Adapting Landmark-Aware Image Models for Facial Expression Recognition in Videos](https://mlsanity.com/paper/2312.05447/) (2023)
- [Accelerating Streaming Video Large Language Models via Hierarchical Token Compression](https://mlsanity.com/paper/2512.00891/) (2025)

## Built on by (most cited first, 6 of 6)

- [RP-OPSD: Resolution-Privileged On-Policy Self-Distillation for Multimodal Large Language Models](https://mlsanity.com/paper/2607.24447/) (2026)
- [GMoT: Gated Motion-Aware Tokenization for Fine-Grained Micro-Gesture Video Reasoning with Multimodal LLMs](https://mlsanity.com/paper/2607.16322/) (2026)
- [PCA: Persistence-Aware Compression and Aggregation for Fast Video Large Language Models](https://mlsanity.com/paper/2607.22726/) (2026)
- [DSPO: Diversity-aware Subjective Policy Optimization for Robust Emotional Reasoning](https://mlsanity.com/paper/2609.36775/) (2026)
- [Learning to Reason with Compressed Context: Ground-Truth-Free Adaptation of OmniLLMs via Self-Distillation](https://mlsanity.com/paper/2609.39953/) (2026)
- [SpatialOPSD: Self-Distilling Spatial Intelligence from Verified Coding Agent Traces](https://mlsanity.com/paper/2610.11366/) (2026)

## Abstract (arXiv)

Reinforcement learning for multimodal large language models (MLLMs) is often hindered by severe reward sparsity in complex reasoning tasks. This challenge is particularly pronounced in human-centered scenarios involving states, emotions, intentions, and behaviors, where heterogeneous multimodal signals and subjective human factors make high-quality chain-of-thought (CoT) annotations expensive and difficult to obtain. Although many multimodal datasets provide expert-annotated ground-truth labels, directly using these labels for supervised fine-tuning may encourage shortcut learning in multimodal perception and provides limited transparency for safety-critical human--AI interaction. To address these limitations, we propose OmniOPSD, a Rationale-Privileged On-Policy Self-Distillation framework that uses frontier-generated rationales as teacher-side privileged evidence rather than student imitation targets. OmniOPSD uses frontier-generated evidence-aware rationales only as training-time privileged evidence context for a local teacher. The student samples its own rollout from the original multimodal input, while the rationale-privileged teacher scores the same tokens and provides dense token-level supervision. Thus, the student learns on its own trajectory distribution without directly imitating frontier-model completions, and inference requires no labels, rationales, CoT annotations, or closed-source model access. Experiments on MER-UniBench show that OmniOPSD achieves state-of-the-art performance with an average score of 84.19, and ablations further support the value of rationale-privileged teacher guidance.