| title | Consistency-Driven Co-Evolution for Self-Supervised Cross-Representation Learning |
|---|---|
| authors | Xuehang Guo, Pengyuan Li, Tom Hope, Tirthankar Ghosal, Manling Li, Qingyun Wang |
| date | 2026-08-05 |
| venue | arXiv (preprint, not peer reviewed) |
| id | arXiv:2608.04926 |
| affiliation | AI4Research |
| upvotes | 4 (Hugging Face Daily Papers, 2026-11-01) |
| links | abs pdf hf |
| abstract | 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 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/. |
cite
@misc{guo2026consistency,
title={Consistency-Driven Co-Evolution for Self-Supervised Cross-Representation Learning},
author={Xuehang Guo and Pengyuan Li and Tom Hope and Tirthankar Ghosal and Manling Li and Qingyun Wang},
year={2026},
eprint={2608.04926},
archivePrefix={arXiv},
url={https://arxiv.org/abs/2608.04926}
}