iGVLM: Dynamic Instruction-Guided Vision Encoding for Question-Aware Multimodal Understanding
This addresses a representation bottleneck in multimodal AI for tasks requiring task-specific visual cues, offering a plug-and-play solution, though it is incremental as it builds on existing LVLM architectures.
The paper tackles the problem of static, instruction-agnostic vision encoders in Large Vision-Language Models, which hinder fine-grained reasoning, by proposing iGVLM, a framework that uses a decoupled dual-branch architecture with Adaptive Layer Normalization to enable instruction-guided visual modulation, resulting in enhanced instruction sensitivity across diverse language backbones.
Despite the success of Large Vision--Language Models (LVLMs), most existing architectures suffer from a representation bottleneck: they rely on static, instruction-agnostic vision encoders whose visual representations are utilized in an invariant manner across different textual tasks. This rigidity hinders fine-grained reasoning where task-specific visual cues are critical. To address this issue, we propose iGVLM, a general framework for instruction-guided visual modulation. iGVLM introduces a decoupled dual-branch architecture: a frozen representation branch that preserves task-agnostic visual representations learned during pre-training, and a dynamic conditioning branch that performs affine feature modulation via Adaptive Layer Normalization (AdaLN). This design enables a smooth transition from general-purpose perception to instruction-aware reasoning while maintaining the structural integrity and stability of pre-trained visual priors. Beyond standard benchmarks, we introduce MM4, a controlled diagnostic probe for quantifying logical consistency under multi-query, multi-instruction settings. Extensive results show that iGVLM consistently enhances instruction sensitivity across diverse language backbones, offering a plug-and-play paradigm for bridging passive perception and active reasoning.