Learning More from Less: Exploiting Counterfactuals for Data-Efficient Chart Understanding

2026-05-11

AI Summary

Existing vision-language models for chart understanding rely heavily on large-scale synthetic data for supervised fine-tuning, struggling to capture the semantic sensitivity of charts to subtle visual variations. To address this limitation, this work introduces counterfactual reasoning into chart understanding in a systematic manner for the first time, proposing a unified framework that integrates code-driven counterfactual data augmentation, chart-similarity-based high-quality sample selection, and multimodal preference optimization. This approach substantially improves data efficiency, achieving competitive or superior performance compared to current state-of-the-art specialized models across five mainstream chart understanding benchmarks while using only a small amount of training data.

Abstract

Vision-Language Models (VLMs) have demonstrated remarkable progress in chart understanding, largely driven by supervised fine-tuning (SFT) on increasingly large synthetic datasets. However, scaling SFT data alone is inefficient and overlooks a key property of charts: charts are programmatically generated visual artifacts, where small, code-controlled visual changes can induce drastic shifts in semantics and correct answers. Learning this counterfactual sensitivity requires VLMs to discriminate fine-grained visual differences, yet standard SFT treats training instances independently and provides limited supervision to enforce this behavior. To address this, we introduce ChartCF, a data-efficient training framework designed to enhance counterfactual sensitivity. ChartCF consists of: (1) a counterfactual data synthesis pipeline via code modification, (2) a chart similarity-based data selection strategy that filters overly difficult samples for improved training efficiency, and (3) multimodal preference optimization across both textual and visual modalities. Experiments on five benchmarks show that ChartCF achieves superior or comparable performance to strong chart-specific VLMs while using significantly less training data.

Problem

• chart understanding
• counterfactual sensitivity
• data efficiency
• vision-language models
• visual reasoning

Innovation

• counterfactual learning
• data-efficient training
• chart understanding
• vision-language models
• multimodal preference optimization