DPC-VQA: Decoupling Quality Perception and Residual Calibration for Video Quality Assessment

Xinyue Li; Shubo Xu; Zhichao Zhang; Zhaolin Cai; Yitong Chen; Guangtao Zhai

cs.CVApr 14, 2026

DPC-VQA: Decoupling Quality Perception and Residual Calibration for Video Quality Assessment

Authors: Xinyue Li, Shubo Xu, Zhichao Zhang, Zhaolin Cai, Yitong Chen, Guangtao Zhai

Organizations: Shanghai Jiao Tong University, Shanghai, China · Baidu Inc., Beijing, China · Xinjiang University, Urumqi, China

Abstract

Recent multimodal large language models (MLLMs) have shown promising performance on video quality assessment (VQA) tasks. However, adapting them to new scenarios remains expensive due to large-scale retraining and costly mean opinion score (MOS) annotations. In this paper, we argue that a pretrained MLLM already provides a useful perceptual prior for VQA, and that the main challenge is to efficiently calibrate this prior to the target MOS space. Based on this insight, we propose DPC-VQA, a decoupling perception and calibration framework for video quality assessment. Specifically, DPC-VQA uses a frozen MLLM to provide a base quality estimate and perceptual prior, and employs a lightweight calibration branch to predict a residual correction for target-scenario adaptation. This design avoids costly end-to-end retraining while maintaining reliable performance with lower training and data costs. Extensive experiments on both user-generated content (UGC) and AI-generated content (AIGC) benchmarks show that DPC-VQA achieves competitive performance against representative baselines, while using less than 2% of the trainable parameters of conventional MLLM-based VQA methods and remaining effective with only 20% of MOS labels. The code will be released upon publication.

Figures & tables

Figure 1. Comparison between existing MLLM-based VQA pipelines and the proposed DPC-VQA framework. Top left: existing methods typically predict quality scores by training or fine-tuning an MLLM with a regressor. Bottom left: when the target scenario changes, they often require scenario-specific retraining with costly MOS supervision. Right: DPC-VQA instead preserves the pretrained MLLM as a frozen quality prior and adapts it through lightweight residual calibration.

Figure 2. Motivation for decoupling quality perception and residual calibration. (a) Frozen MLLM quality predictions are already correlated with MOS, indicating a useful base quality prior. (b) The residual error shows a clear dependency on the base score, suggesting that the remaining discrepancy is structured and learnable. (c) Residual errors are concentrated around zero, implying that most samples require only limited correction. These results motivate calibrating a perceptually grounded base score rather than relearning video quality prediction end-to-end. All results are computed on the KoNViD-1k dataset for analysis.

Figure 3. Overview of the proposed DPC-VQA framework. A frozen MLLM first produces a base quality estimate and perceptual representations as a quality prior. A lightweight base-aware residual calibration branch then predicts token-wise corrections conditioned on the base estimate and aggregates them into a residual adjustment. The final quality score is obtained by combining the base estimate with the aggregated residual, enabling efficient adaptation under limited MOS supervision.

supervised
Methods	KoNViD-1k		YouTube-UGC		LIVE-VQC		T2VQA		Human-AGVQA
Methods	SRCC	PLCC	SRCC	PLCC	SRCC	PLCC	SRCC	PLCC	SRCC	PLCC
BRISQUE (TIP, 2012) ( Mittal et al., 2012 )	0.634	0.615	0.472	0.447	0.459	0.451	0.505	0.507	0.438	0.446
TLVQM (TIP, 2019) ( Korhonen, 2019 )	0.773	0.769	0.669	0.659	0.658	0.639	0.610	0.613	0.582	0.584
VSFA (ACMMM, 2019) ( Li et al., 2019 )	0.794	0.799	0.787	0.789	0.718	0.771	0.704	0.708	0.618	0.633
RAPIQUE (OJSP, 2021) ( Tu et al., 2021 )	0.811	0.819	0.774	0.781	0.740	0.764	0.682	0.703	0.632	0.648

Table 1. Comparison with representative supervised and few-shot methods on five VQA benchmarks spanning both UGC and AIGC scenarios. Methods marked with ∗ are designed for AIGC-VQA and are therefore reported only on AIGC benchmarks.

Method	KoNViD-1k		YouTube-UGC
Method	SRCC	PLCC	SRCC	PLCC
Base only ( $\hat{y}=q_{b}$ )	0.774	0.786	0.760	0.753
Direct regression	0.836	0.844	0.821	0.818
Score-conditioned regression	0.868	0.876	0.849	0.855
Residual calibration (ours)	0.883	0.891	0.868	0.871

Table 2. Comparison between residual calibration and direct regression on KoNViD-1k and YouTube-UGC.

Inputs	KoNViD-1k		YouTube-UGC
Inputs	SRCC	PLCC	SRCC	PLCC
$H_{\mathrm{vis}}^{L}$ only	0.834	0.842	0.814	0.820
$H_{\mathrm{aux}}$ only	0.809	0.817	0.787	0.794
$H_{\mathrm{vis}}^{L}+H_{\mathrm{aux}}$	0.854	0.862	0.836	0.841
$H_{\mathrm{vis}}^{L}+H_{\mathrm{aux}}+q_{b}$	0.875	0.883	0.858	0.862
$H_{\mathrm{vis}}^{L}+H_{\mathrm{aux}}+q_{b}+u_{b}$	0.883	0.891	0.868	0.871

Table 3. Ablation on different information sources used for calibration on KoNViD-1k and YouTube-UGC.

Aggregation strategy	KoNViD-1k		YouTube-UGC
Aggregation strategy	SRCC	PLCC	SRCC	PLCC
Mean pooling	0.859	0.867	0.842	0.847
Self-attention aggregation	0.874	0.882	0.859	0.865
Calibration queries (ours)	0.883	0.891	0.868	0.871

Table 4. Ablation on different evidence aggregation strategies on KoNViD-1k and YouTube-UGC.

$M$	KoNViD-1k		YouTube-UGC
$M$	SRCC	PLCC	SRCC	PLCC
1	0.861	0.869	0.846	0.850
4	0.874	0.882	0.858	0.863
8	0.883	0.891	0.868	0.871
16	0.879	0.887	0.864	0.868

Table 5. Sensitivity analysis on the number of calibration queries

M

on KoNViD-1k and YouTube-UGC.

$\lambda_{\mathrm{res}}$	KoNViD-1k		YouTube-UGC
$\lambda_{\mathrm{res}}$	SRCC	PLCC	SRCC	PLCC
0	0.869	0.877	0.853	0.857
0.01	0.875	0.881	0.859	0.862
0.05	0.883	0.891	0.868	0.871
0.1	0.880	0.888	0.864	0.868

Table 6. Effect of the residual regularization coefficient

\lambda_{\mathrm{res}}

on KoNViD-1k and YouTube-UGC.

$\alpha$	KoNViD-1k		YouTube-UGC
$\alpha$	SRCC	PLCC	SRCC	PLCC
0.1	0.871	0.879	0.856	0.860
0.2	0.883	0.891	0.868	0.871
0.3	0.876	0.884	0.860	0.865
0.5	0.867	0.875	0.851	0.856

Table 7. Effect of the residual bound

\alpha

on KoNViD-1k and YouTube-UGC.

Figure 4. Data efficiency comparison under different training ratios on KoNViD-1k and YouTube-UGC.

Explore similar work

VersusQ: Pairwise Margin Reasoning for Generalizable Video Quality Assessment
May 20, 2026Shibei Meng, Binxin Yang, Yuan Liu +4Pairwise Comparison Learning to Rank
The Chinese University of Hong Kong · Beijing Normal University · WeChat Vision, Tencent Inc. +2
LLaVA-Assessor: Building the Foundation LMM For Visual Quality Assessment
Aug 11, 2026Ziheng Jia, Zicheng Zhang, Jiaying Qian +2Image Quality Assessment Video Quality Assessment
Shanghai Artificial Intelligence Laboratory, Shanghai 200032, China
Analytic Score Optimization for Multi Dimension Video Quality Assessment
Feb 18, 2026Boda Lin, Yongjie Zhu, Wenyu Qin +2Video Quality Assessment
Kling Team, Kuaishou Technology, Beijing, China.

May 20, 2026J/K move · Enter open · S save

VersusQ: Pairwise Margin Reasoning for Generalizable Video Quality Assessment

Shibei Meng, Binxin Yang, Yuan Liu, Jiexuan Zhang+3

The Chinese University of Hong Kong · Beijing Normal University · WeChat Vision, Tencent Inc.+2

Pairwise Comparison Learning to Rank Pairwise Ranking Video Quality Assessment VLM Reasoning

Open paper AskPDF arXiv

Large Multimodal Models (LMMs) have shown promise for video quality assessment, but most methods still predict an absolute score for each video. Such pointwise supervision often mixes perceptual quality with dataset-specific calibration, including annotation protocols, rating habits, and score distributions. As a result, the learned scoring rule may work well within a benchmark but transfer poorly across unseen domains. We argue that relative comparisons alleviate the absolute-scale calibration bias by focusing purely on perceptual differences rather than dataset-specific rating habits. Consequently, we propose VersusQ, a pairwise margin reasoning framework driven entirely by direct comparisons. Specifically, VersusQ performs LMM-based comparison between two videos, reasons about their visual and temporal quality differences, and predicts a signed continuous margin that captures both the preferred choice and the degree of difference. Furthermore, to align interpretable comparison rationales with fine-grained numerical differences, we introduce Margin-Coupled GRPO, which jointly optimizes rollout-based relational reasoning and continuous margin regression. Extensive experiments on multiple public VQA benchmarks demonstrate that VersusQ achieves state-of-the-art performance, strong cross-domain generalization, and reliable fine-grained ranking under heterogeneous evaluation scenarios.

Abstract

Explore similar work

VersusQ: Pairwise Margin Reasoning for Generalizable Video Quality Assessment

LLaVA-Assessor: Building the Foundation LMM For Visual Quality Assessment

Analytic Score Optimization for Multi Dimension Video Quality Assessment