Saved in:
Bibliographic Details
Main Authors: Du, Yao, Song, Shanshan, Li, Xiaomeng
Format: Preprint
Published: 2026
Subjects:
Online Access:https://arxiv.org/abs/2605.01402
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866913109346615296
author Du, Yao
Song, Shanshan
Li, Xiaomeng
author_facet Du, Yao
Song, Shanshan
Li, Xiaomeng
contents Multimodal large language models (MLLMs) struggle with numerical regression under long-tailed target distributions. Token-level supervised fine-tuning (SFT) and point-wise regression rewards bias learning toward high-density regions, leading to regression-to-the-mean behavior and poor tail performance. We identify the lack of cross-sample relational supervision as a key limitation of existing MLLM training paradigms. To address it, we propose a distribution-aware reinforcement learning framework based on Group Relative Policy Optimization, which introduces batch-level comparison-based supervision via the Concordance Correlation Coefficient-based reward to align predicted and ground-truth distributions in terms of correlation, scale, and mean. The framework is plug-and-play, requiring no architectural modification. Experiments on a unified suite of long-tailed regression benchmarks show consistent improvements over SFT and existing MLLM regression methods, with particularly strong gains in medium- and few-shot regimes.
format Preprint
id arxiv_https___arxiv_org_abs_2605_01402
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle Injecting Distributional Awareness into MLLMs via Reinforcement Learning for Deep Imbalanced Regression
Du, Yao
Song, Shanshan
Li, Xiaomeng
Computation and Language
Computer Vision and Pattern Recognition
Machine Learning
Multimodal large language models (MLLMs) struggle with numerical regression under long-tailed target distributions. Token-level supervised fine-tuning (SFT) and point-wise regression rewards bias learning toward high-density regions, leading to regression-to-the-mean behavior and poor tail performance. We identify the lack of cross-sample relational supervision as a key limitation of existing MLLM training paradigms. To address it, we propose a distribution-aware reinforcement learning framework based on Group Relative Policy Optimization, which introduces batch-level comparison-based supervision via the Concordance Correlation Coefficient-based reward to align predicted and ground-truth distributions in terms of correlation, scale, and mean. The framework is plug-and-play, requiring no architectural modification. Experiments on a unified suite of long-tailed regression benchmarks show consistent improvements over SFT and existing MLLM regression methods, with particularly strong gains in medium- and few-shot regimes.
title Injecting Distributional Awareness into MLLMs via Reinforcement Learning for Deep Imbalanced Regression
topic Computation and Language
Computer Vision and Pattern Recognition
Machine Learning
url https://arxiv.org/abs/2605.01402