| title | Task-Conditional Flow Matching for Balanced Multilingual Text Embedding Adaptation |
|---|---|
| authors | Tirth Bhatt, Naren Kumar S, Mayank Singh |
| date | 2026-08-06 |
| venue | arXiv (preprint, not peer reviewed) |
| id | arXiv:2608.05785 |
| affiliation | Lingo Research Group |
| upvotes | 3 (Hugging Face Daily Papers, 2026-11-01) |
| links | abs pdf hf |
| abstract | Multilingual text embedding models are commonly adapted using a single training objective across diverse tasks, despite different tasks requiring fundamentally different optimization strategies. We introduce Task-Conditional Flow Matching (TCFM), a multilingual embedding adaptation framework that selectively applies Flow Matching to translation tasks while optimizing retrieval, classification, and pair-classification tasks with objectives better aligned to their learning dynamics. TCFM further combines teacher-guided representation preservation with a three-stage curriculum to enable stable adaptation. Evaluated on the Indic Massive Text Embedding Benchmark, TCFM establishes a new state-of-the-art, consistently improving embedding quality across a diverse set of multilingual tasks and generalizing across embedding model families. We will publicly release the codebase and datasets upon acceptance of the paper. |
cite
@misc{bhatt2026task,
title={Task-Conditional Flow Matching for Balanced Multilingual Text Embedding Adaptation},
author={Tirth Bhatt and Naren Kumar S and Mayank Singh},
year={2026},
eprint={2608.05785},
archivePrefix={arXiv},
url={https://arxiv.org/abs/2608.05785}
}