Problem
Research questions and friction points this paper is trying to address.
chart understanding
structured data recovery
factual summary
data extraction
natural-language generation
ChartLens: A Dual-Branch Framework for Chart Data Correction and Factual Summary Refinement
🤖 AI Summary
This work addresses the challenge of ensuring factual consistency in structured data recovery and natural language summarization from chart images by proposing ChartLens, a dual-branch framework. The approach innovatively integrates a structure-aware CSV verification and correction module with a text-preserving guided summarization module. By leveraging OCR-assisted evidence alignment, model adaptation, and a correction-based generation strategy, the framework jointly enhances both data extraction accuracy and summary faithfulness. Evaluated on the Track 2 test set of the DataMFM Challenge, the proposed method achieves a top-ranked overall score of 69.10, significantly improving data reliability and factual consistency in chart understanding.
📝 Abstract
In this report, we present our champion solution for the DataMFM Challenge Track 2: Chart Understanding. This track requires models to recover structured chart data and generate faithful natural-language summaries from chart images. To address the complementary requirements of accurate data extraction and factual narration, we propose ChartLens, a dual-branch framework for chart data correction and summary refinement. ChartLens consists of two key modules: Structure-Aware CSV Verification and Correction (SAVC) and Text-Retention-Guided Summary Refinement (TRSR). SAVC improves the reliability of structured data extraction through verification and correction, while TRSR enhances summary generation by preserving critical textual and numerical evidence from charts. By combining model adaptation, correction-based generation, and OCR-assisted evidence grounding, ChartLens improves both structured data recovery and summary factuality. On the test set, our final system achieves an overall score of 69.10 and ranks first in Track 2, demonstrating its effectiveness for accurate chart understanding. Our code will be released at: https://github.com/iLearn-Lab/CVPRW26-ChartLens.
Innovation
Methods, ideas, or system contributions that make the work stand out.
dual-branch framework
structured data correction
factual summary refinement
OCR-assisted grounding
chart understanding
🔎 Similar Papers
No similar papers found.
