# Through the Lens of Character: Resolving Modality-Role Interference in Multimodal Role-Playing Agent

Source: [Through the Lens of Character: Resolving Modality-Role Interference in Multimodal Role-Playing Agent](<https://www.sota2.com/research/paper/6a02b469dbc836b87d824b21>)

Collection index: [llms.txt](<https://www.sota2.com/research/llms.txt>)

Data updated: 2026-05-13T05:00:12.693Z

Authors: Yihong Tang, Kehai Chen, Xuefeng Bai, Min Zhang

Published: 2026-05-10T09:40:43.000Z

## Abstract

The advancement of Multimodal Large Language Models (MLLMs) has expanded Role-Playing Agents (RPAs) into visually grounded environments. However, human vision is inherently subjective and identity-driven, whereas existing MLLMs extract objective, character-agnostic features for general tasks. In RPAs, this generic visual noise overpowers fragile character traits, causing Modality-Role Interference (MRI), where agents struggle to integrate visual grounding and character consistency. To address this, we introduce the training-free Character-Aware Visual Intervention (CAVI) framework, enabling agents to perceive the world through the lens of character. CAVI systematically targets MRI: macroscopically, Character-Guided Token Pruning (CTP) restricts the visual receptive field to role-relevant entities; microscopically, Orthogonal Feature Modulation (OFM) projects tokens onto a character-context subspace to extract aligned facts; and during decoding, Modality-Adaptive Role Steering (MARS) dynamically optimizes steering intensity based on visual reliance. Extensive experiments show CAVI effectively alleviates MRI, significantly enhancing character-consistent multimodal interactions.

## Recorded benchmark results

A research result applies to its stated dataset, split, metric and evaluation configuration. Compare like-for-like settings, preserve the metric's unit and direction, and distinguish paper-reported results from independent reproduction. The best result recorded here is not a claim of exhaustive, current global coverage.

- [Multi-modal Role-playing on MMRole in-domain](<https://www.sota2.com/research/sota/multi-modal-role-playing-on-mmrole-in-domain.md>): Multi-modal Role-playing; MMRole in-domain; IA: 1.135.

- [Multimodal Role-playing on MMRole (out-of-domain)](<https://www.sota2.com/research/sota/multimodal-role-playing-on-mmrole-out-of-domain.md>): Multimodal Role-playing; MMRole (out-of-domain); IA: 1.111.

## Original sources and implementation

[Original paper](<https://arxiv.org/abs/2605.09443>)

[PDF](<https://arxiv.org/pdf/2605.09443.pdf>)