Consistency-Driven Co-Evolution for Self-Supervised Cross-Representation Learning

Source

• Source name: arXiv
• Source kind: paper
• Authors / organization: Xuehang Guo, Pengyuan Li, Tom Hope, Tirthankar Ghosal, Manling Li, Qingyun Wang
• Venue: arXiv: cs.LG
• Published: 2026-08-05
• Collected: 2026-08-06
• URL: http://arxiv.org/abs/2608.04926v1
• PDF: https://arxiv.org/pdf/2608.04926v1
• DOI: UNKNOWN

Plain-Language Brief

The paper introduces CoCoEvolve, a method aimed at improving the consistency between chart, table, and code representations in AI systems. The method optimizes models based on agreement between different representations without requiring additional annotations. It uses co-evolution during training and test-time co-optimization for evaluation.

What the Document Says

• The paper presents CoCoEvolve, a method to improve consistency across chart, table, and code representations in AI systems.
• Instead of treating cross-representation mapping as one-to-many, it defines explicit one-to-one correspondences and optimizes models based on agreement between representations without additional annotations.
• During training, CoCoEvolve@Train performs co-evolution across the chart-table-code cycle.
• At inference time, CoCoEvolve@Test applies the same consistency objective for test-time co-optimization.
• A CoCoEvolve@Eval evaluation suite is presented for all six cross-representation tasks.
• The method improves performance in both training and test settings across four benchmarks.

Key Concepts Mentioned

• AI systems
• Cross-representation understanding
• Chart images
• Tabular data
• Visualization code
• Co-evolution
• Consistency objective
• Evaluation suite

Claims Made by the Document

• CoCoEvolve improves consistency across chart, table, and code representations.
• It performs better in both training and test settings across four benchmarks.

Evidence Used

• The paper describes the methodology and results of CoCoEvolve.

Methods / Approach

• Defining explicit one-to-one correspondences
• Co-evolution during training
• Test-time co-optimization
• Evaluation suite for cross-representation tasks

Limitations Stated by the Authors

Not stated in source

Limitations Noted by the Agent

Agent interpretation — not from the source.

The method may not perform well on datasets outside the four benchmarks mentioned.

Open Questions

• What are the specific algorithms used in CoCoEvolve?
• How does CoCoEvolve handle inconsistencies that can't be resolved by one-to-one correspondences?

Notes for Future Q&A

May include interpretation. Labeled where it does.

The paper provides a high-level overview of the method without delving into specific algorithmic details.

Raw Abstract / Excerpt

As chart images, tabular data, and visualization code play increasingly important roles across diverse domains, cross-representation understanding across these modalities poses fundamental challenges for AI systems: the relationships across representations are inherently \textit{one-to-many}, supervision is ambiguous and costly, and model optimization lacks a principled signal that is both direction-adaptive and representation-generalizable beyond task-specific objectives. We introduce CoCoEvolve to improve consistency across chart, table, and code representations. Instead of treating cross-representation mapping as a one-to-many problem, we define explicit one-to-one correspondences and optimize models using agreement between representations, without additional annotations. During training, CoCoEvolve@Train performs co-evolution across the chart-table-code cycle, while CoCoEvolve@Test applies the same consistency objective at inference time for test-time co-optimization. We also present CoCoEvolve@Eval, an evaluation suite covering all six cross-representation tasks. Across four benchmarks, CoCoEvolve improves performance in both training-time and test-time settings. Our project page: https://xhguo7.github.io/CoCoEvolve/.