Researchers have introduced ChartREG++, a new benchmark designed to improve and evaluate the grounding of referring expressions in charts. This benchmark addresses limitations in existing datasets by supporting multiple localization forms, handling multi-instance targets, incorporating diverse grounding cues beyond simple text, and covering a wider array of chart types. The paper also presents a code-driven synthesis pipeline to generate pixel-accurate instance masks for training an instance segmentation model, which, when integrated into a multimodal grounding framework, demonstrates superior performance on the new benchmark and generalizes to other chart-grounding tasks. AI
Summary written by gemini-2.5-flash-lite from 1 source. How we write summaries →
IMPACT Establishes a new standard for evaluating multimodal models on chart understanding, potentially driving improvements in visual grounding and reasoning capabilities.
RANK_REASON The cluster contains an academic paper introducing a new benchmark for a specific AI task. [lever_c_demoted from research: ic=1 ai=1.0]