SOURCE-LINKED INTELLIGENCE

DICS: Exploring Data Intrinsic Consistency for Visual Instruction Selection

arXiv · AI, language, vision and robotics · article · Aug 31, 2026 · UTC

Visual instruction tuning is crucial for advancing the vision-language alignment and instruction-following capabilities of Vision-Language Models (VLMs). However, identifying optimal subsets under a fixed ratio constraint from rapidly expanding datasets remains a significant bottleneck. While existing methods largely depend on distribution diversity or heuristic filtering, they often overlook the internal coherence within individual samples. To bridge this gap, we propose Data Intrinsic Consistency (DIC), a self-scoring metric designed to quantify the sample-level inter-component consistency.

Read original source ↗ Open in workspace

recordType

paper

region

Global