CycleChart: A Unified Consistency-Based Learning Framework for Bidirectional Chart Understanding and Generation

TL;DR

<3-5 sentence high-level summary> CycleChart presents a unified, consistency-based framework that jointly learns chart generation, schema parsing, data extraction, and chart question answering by enforcing a closed generate–parse–reason cycle. It introduces CycleChart-Bench, a lifecycle-aligned dataset built atop nvBench 2.0 with facet-augmented charts and aligned supervision across tasks. The framework demonstrates substantial gains across NL2Chart, chart parsing, data parsing, and ChartQA, and generalizes well to external benchmarks, showing the value of bidirectional semantic consistency for chart understanding. This work advances general chart reasoning and offers a rigorous benchmark and training paradigm for future multimodal chart analysis systems.

Abstract

Current chart-specific tasks, such as chart question answering, chart parsing, and chart generation, are typically studied in isolation, preventing models from learning the shared semantics that link chart generation and interpretation. We introduce CycleChart, a consistency-based learning framework for bidirectional chart understanding and generation. CycleChart adopts a schema-centric formulation as a common interface across tasks. We construct a consistent multi-task dataset, where each chart sample includes aligned annotations for schema prediction, data parsing, and question answering. To learn cross-directional chart semantics, CycleChart introduces a generate-parse consistency objective: the model generates a chart schema from a table and a textual query, then learns to recover the schema and data from the generated chart, enforcing semantic alignment across directions. CycleChart achieves strong results on chart generation, chart parsing, and chart question answering, demonstrating improved cross-task generalization and marking a step toward more general chart understanding models.

Figures (4)

• Figure 1: Overview of the chart creation pipeline, adapted from GoG. Charts are produced by transforming a source table, mapping it to a schema, and rendering an image. NL2Chart follows the creation path, Chart Parsing inverts it, and ChartQA reasons over the resulting chart.
• Figure 2: CycleChart-Bench construction. nvBench-2.0 queries produce single-view charts, while our facet-oriented queries generate composite charts. Both chart types supply transformed tables and QA pairs for aligned parsing and reasoning supervision.
• Figure 3: Overview of the CycleChart training framework. Given a source table and NL query, the model generates a chart specification (NL2Chart), which is executed to obtain a chart image and its visualization-level table. The model then parses the generated chart into a schema and table (Chart Parsing) and answers NL questions about the chart (ChartQA). All tasks use cross-entropy supervision, forming a closed generate--parse--reason cycle that enforces consistency.
• Figure 4: Impact of training steps across benchmarks.