# DSPO: Diversity-aware Subjective Policy Optimization for Robust Emotional Reasoning

arXiv: https://arxiv.org/abs/2609.36775 · HTML: https://mlsanity.com/paper/2609.36775/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 45 of its 58 references that are in the map.
- First posted on arXiv 29 Sep 2026.

## Placed in

- LLM reasoning

## Builds on (most cited first, 30 of 45)

- [DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Models](https://mlsanity.com/paper/2402.03300/) (2024)
- [Qwen2.5-VL Technical Report](https://mlsanity.com/paper/2502.13923/) (2025)
- [Improved Baselines with Visual Instruction Tuning](https://mlsanity.com/paper/2310.03744/) (2023)
- [DAPO: An Open-Source LLM Reinforcement Learning System at Scale](https://mlsanity.com/paper/2503.14476/) (2025)
- [Ferret: Refer and Ground Anything Anywhere at Any Granularity](https://mlsanity.com/paper/2310.07704/) (2023)
- [OPERA: Alleviating Hallucination in Multi-Modal Large Language Models via Over-Trust Penalty and Retrospection-Allocation](https://mlsanity.com/paper/2311.17911/) (2023)
- [Mitigating Hallucination in Large Multi-Modal Models via Robust Instruction Tuning](https://mlsanity.com/paper/2306.14565/) (2023)
- [Monkey: Image Resolution and Text Label Are Important Things for Large Multi-modal Models](https://mlsanity.com/paper/2311.06607/) (2023)
- [Analyzing and Mitigating Object Hallucination in Large Vision-Language Models](https://mlsanity.com/paper/2310.00754/) (2023)
- [MiMo-V2-Flash Technical Report](https://mlsanity.com/paper/2601.02780/) (2026)
- [Oryx MLLM: On-Demand Spatial-Temporal Understanding at Arbitrary Resolution](https://mlsanity.com/paper/2409.12961/) (2024)
- [Emotion-LLaMA: Multimodal Emotion Recognition and Reasoning with Instruction Tuning](https://mlsanity.com/paper/2406.11161/) (2024)
- [R1-Omni: Explainable Omni-Multimodal Emotion Recognition with Reinforcement Learning](https://mlsanity.com/paper/2503.05379/) (2025)
- [Evaluation and Analysis of Hallucination in Large Vision-Language Models](https://mlsanity.com/paper/2308.15126/) (2023)
- [Thinking With Videos: Multimodal Tool-Augmented Reinforcement Learning for Long Video Reasoning](https://mlsanity.com/paper/2508.04416/) (2025)
- [MLLM can see? Dynamic Correction Decoding for Hallucination Mitigation](https://mlsanity.com/paper/2410.11779/) (2024)
- [EmoSet: A Large-scale Visual Emotion Dataset with Rich Attributes](https://mlsanity.com/paper/2307.07961/) (2023)
- [EmoBank: Studying the Impact of Annotation Perspective and Representation Format on Dimensional Emotion Analysis](https://mlsanity.com/paper/2205.01996/) (2022)
- [ThinkMorph: Emergent Properties in Multimodal Interleaved Chain-of-Thought Reasoning](https://mlsanity.com/paper/2510.27492/) (2025)
- [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)
- [VCoder: Versatile Vision Encoders for Multimodal Large Language Models](https://mlsanity.com/paper/2312.14233/) (2023)
- [Contemplating Visual Emotions: Understanding and Overcoming Dataset Bias](https://mlsanity.com/paper/1808.02212/) (2018)
- [Context De-confounded Emotion Recognition](https://mlsanity.com/paper/2303.11921/) (2023)
- [Multi-Attention Network for Compressed Video Referring Object Segmentation](https://mlsanity.com/paper/2207.12622/) (2022)
- [Video-STAR: Reinforcing Open-Vocabulary Action Recognition with Tools](https://mlsanity.com/paper/2510.08480/) (2025)
- [Graph Mixture of Experts and Memory-augmented Routers for Multivariate Time Series Anomaly Detection](https://mlsanity.com/paper/2412.19108/) (2024)
- [Towards Efficient Partially Relevant Video Retrieval with Active Moment Discovering](https://mlsanity.com/paper/2504.10920/) (2025)
- [MME-Emotion: A Holistic Evaluation Benchmark for Emotional Intelligence in Multimodal Large Language Models](https://mlsanity.com/paper/2508.09210/) (2025)
- [EmotionThinker: Prosody-Aware Reinforcement Learning for Explainable Speech Emotion Reasoning](https://mlsanity.com/paper/2601.15668/) (2026)

## Abstract (arXiv)

Reinforcement Learning has significantly advanced the complex reasoning capabilities of MLLMs. However, prevailing RL algorithms suffer a severe failure in emotion reasoning tasks. These methods heavily rely on deterministic hard-label supervision and point-wise isolated evaluation, creating a fundamental gap with the inherently subjective and continuously distributed nature of human emotions. Furthermore, unlike explicit physical objects, emotional states are deeply implicit within visual cues. This abstract nature exacerbates visual hallucinations in MLLMs, leading to plausible yet ungrounded emotional evidence. To address these limitations, we propose Diversity-Aware Subjective Policy Optimization (DSPO), a reinforcement learning framework that jointly promotes subjective affective coverage and visual grounding. First, we construct a context-grounded emotional distribution prior in the VAD space by combining the lexical prior of the annotated emotion with image-specific contextual information. Based on this prior, we introduce a Distribution-Aligned Emotional Diversity Reward (DEDR), which measures the leave-one-out marginal contribution of each candidate emotion within a rollout. DEDR rewards candidates whose inclusion brings the predicted affective set closer to the context-grounded prior, thereby preserving plausible subjective interpretations without encouraging unconstrained dispersion. We further develop Counterfactual Visual Intervention Gating (CVIG), which masks the visual region highlighted in the reasoning process and uses the resulting candidate-wise probability changes to reduce the weights of interpretations unsupported by visual evidence. Extensive experiments demonstrate that DSPO achieves state-of-the-art performance across multiple public benchmarks, especially on the cross-domain performance, i.e., improving +10.8% on average cross-domain accuracy than EMO-R3.