# What Can Low Resource Languages Learn From Each Other?

arXiv: https://arxiv.org/abs/2608.27753 · HTML: https://mlsanity.com/paper/2608.27753/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Placed in Document understanding and multimodal RAG, one of 16 directions in Multimodal LLMs / vision-language models.
- Builds on 18 of its 25 references that are in the map.
- First posted on arXiv 27 Aug 2026.

## Placed in

- Multimodal LLMs / vision-language models / [Document understanding and multimodal RAG](https://mlsanity.com/mllm/document-understanding-and-multimodal-rag/)

## Builds on (most cited first, 18 of 18)

- [Similarity of Neural Network Representations Revisited](https://mlsanity.com/paper/1905.00414/) (2019)
- [Model soups: averaging weights of multiple fine-tuned models improves accuracy without increasing inference time](https://mlsanity.com/paper/2203.05482/) (2022)
- [Editing Models with Task Arithmetic](https://mlsanity.com/paper/2212.04089/) (2022)
- [TIES-Merging: Resolving Interference When Merging Models](https://mlsanity.com/paper/2306.01708/) (2023)
- [Dataless Knowledge Fusion by Merging Weights of Language Models](https://mlsanity.com/paper/2212.09849/) (2022)
- [TrOCR: Transformer-based Optical Character Recognition with Pre-trained Models](https://mlsanity.com/paper/2109.10282/) (2021)
- [Scene Text Recognition with Permuted Autoregressive Sequence Models](https://mlsanity.com/paper/2207.06966/) (2022)
- [ICDAR2019 Robust Reading Challenge on Multi-lingual Scene Text Detection and Recognition -- RRC-MLT-2019](https://mlsanity.com/paper/1907.00945/) (2019)
- [Model Breadcrumbs: Scaling Multi-Task Model Merging with Sparse Masks](https://mlsanity.com/paper/2312.06795/) (2023)
- [ICDAR 2019 Robust Reading Challenge on Reading Chinese Text on Signboard](https://mlsanity.com/paper/1912.09641/) (2019)
- [SynthTIGER: Synthetic Text Image GEneratoR Towards Better Text Recognition Models](https://mlsanity.com/paper/2107.09313/) (2021)
- [Benchmarking Scene Text Recognition in Devanagari, Telugu and Malayalam](https://mlsanity.com/paper/2104.04437/) (2021)
- [IndicVisionBench: Benchmarking Cultural and Multilingual Understanding in VLMs](https://mlsanity.com/paper/2511.04727/) (2025)
- [Transfer Learning for Scene Text Recognition in Indian Languages](https://mlsanity.com/paper/2201.03180/) (2022)
- [Cross-Lingual Learning in Multilingual Scene Text Recognition](https://mlsanity.com/paper/2312.10806/) (2023)
- [Towards Deployable OCR models for Indic languages](https://mlsanity.com/paper/2205.06740/) (2022)
- [Spanish TrOCR: Leveraging Transfer Learning for Language Adaptation](https://mlsanity.com/paper/2407.06950/) (2024)
- [IndicSTR12: A Dataset for Indic Scene Text Recognition](https://mlsanity.com/paper/2403.08007/) (2024)

## Abstract (arXiv)

Despite the rapid advancement of Vision-Language Models (VLMs), their linguistic reach remains largely confined to high-resource languages, leaving the majority of the world's 7,000+ living languages on the wrong side of a growing digital divide. This disparity is especially pronounced in Optical Character Recognition (OCR), where low-resource scripts lack the massive datasets required for traditional scaling laws. We investigate OCR adaptation in extreme data-scarce regimes (<10K real and <250K synthetic images), demonstrating that conventional fine-tuning strategies often reach a performance ceiling. Our key finding reveals a structural inefficiency in language-specific adaptation: while higher layers of specialized models diverge to capture unique script nuances, the lower layers learn redundant, highly similar features. Motivated by this observation, we propose PSMC (Pre-train, Specialize, Merge, and Co-train), a data-efficient framework that capitalizes on a cross-script "transfer effect". Our approach first derives language-specific experts from a high-resource base model, then employs task arithmetic to fuse these experts into a unified, high-performance multilingual back- bone. Extensive evaluation across 10 Indian scripts (supporting 20+ languages) shows that PSMC achieves a ~2% average improvement in Word Recognition Rate (WRR) over individual specialist models without increasing parameter count. Our results indicate that joint training in the merged latent space facilitates a constructive knowledge transfer that benefits all constituent scripts, providing a scalable pathway for inclusive VLM development. Source code and datasets will be released post publication.