# Soro: A Lightweight Foundation Model and Chatbot for Tajik

arXiv: https://arxiv.org/abs/2605.27379 · HTML: https://mlsanity.com/paper/2605.27379/

## Key facts

- Cited by 2 papers in the map (data through 14 Oct 2026).
- Builds on 20 of its 32 references that are in the map.
- First posted on arXiv 9 Apr 2026.

## Builds on (most cited first, 20 of 20)

- [LoRA: Low-Rank Adaptation of Large Language Models](https://mlsanity.com/paper/2106.09685/) (2021)
- [Measuring Massive Multitask Language Understanding](https://mlsanity.com/paper/2009.03300/) (2020)
- [Finetuned Language Models Are Zero-Shot Learners](https://mlsanity.com/paper/2109.01652/) (2021)
- [GPTQ: Accurate Post-Training Quantization for Generative Pre-trained Transformers](https://mlsanity.com/paper/2210.17323/) (2022)
- [Don't Stop Pretraining: Adapt Language Models to Domains and Tasks](https://mlsanity.com/paper/2004.10964/) (2020)
- [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)
- [Language Models are Super Mario: Absorbing Abilities from Homologous Models as a Free Lunch](https://mlsanity.com/paper/2311.03099/) (2023)
- [Arcee's MergeKit: A Toolkit for Merging Large Language Models](https://mlsanity.com/paper/2403.13257/) (2024)
- [The BigScience ROOTS Corpus: A 1.6TB Composite Multilingual Dataset](https://mlsanity.com/paper/2303.03915/) (2023)
- [Lifting the Curse of Multilinguality by Pre-training Modular Transformers](https://mlsanity.com/paper/2205.06266/) (2022)
- [FineWeb2: One Pipeline to Scale Them All -- Adapting Pre-Training Data Processing to Every Language](https://mlsanity.com/paper/2506.20920/) (2025)
- [FP8 Quantization: The Power of the Exponent](https://mlsanity.com/paper/2208.09225/) (2022)
- [Asymmetry in Low-Rank Adapters of Foundation Models](https://mlsanity.com/paper/2402.16842/) (2024)
- [Breaking the Curse of Multilinguality with Cross-lingual Expert Language Models](https://mlsanity.com/paper/2401.10440/) (2024)
- [TeenyTinyLlama: open-source tiny language models trained in Brazilian Portuguese](https://mlsanity.com/paper/2401.16640/) (2024)
- [VinaLLaMA: LLaMA-based Vietnamese Foundation Model](https://mlsanity.com/paper/2312.11011/) (2023)
- [Rethinking Multilingual Continual Pretraining: Data Mixing for Adapting LLMs Across Languages and Resources](https://mlsanity.com/paper/2504.04152/) (2025)
- [Large Malaysian Language Model Based on Mistral for Enhanced Local Language Understanding](https://mlsanity.com/paper/2401.13565/) (2024)

## Built on by (most cited first, 2 of 2)

- [Benchmarking Parameter-Efficient Fine-Tuning of Large Language Models for Low-Resource Tajik Text Generation with the Tajik Web Corpus](https://mlsanity.com/paper/2605.03742/) (2026)
- [Large Language Models for Low-Resource Languages: A Conceptual Framework for an Electronic Explanatory Dictionary of the Tajik Language](https://mlsanity.com/paper/2608.04186/) (2026)

## Abstract (arXiv)

We present Soro, a family of Tajik-specialized conversational large language models (LLMs) designed for real-world deployment under tight compute and connectivity constraints in Tajikistan. Starting from open-weight Gemma 3 checkpoints, we perform Tajik-only continual pretraining on a curated 1.9-billion-token corpus spanning filtered web text, PDF documents, and curriculum-aligned educational materials, followed by supervised instruction tuning on 40K Tajik teacher-style examples. To enable rigorous evaluation despite the limited coverage of Tajik in standard benchmarks, we introduce a suite of Tajik benchmarks covering general knowledge, linguistic competence, and school- and university entrance-exam domains, and we open-source them on Hugging Face. Across these Tajik benchmarks, Soro substantially outperforms same-size Gemma 3 baselines while retaining strong English performance on standard datasets. We further show that FP8 and INT4 quantization of Soro preserves most Tajik-language gains while reducing memory requirements for edge deployment, supporting an ongoing education-sector pilot and planned scale-out across schools in Tajikistan.