BEIR

university
Activity Feed

AI & ML interests

BEIR (Benchmarking IR) consists of a homogenous benchmark for diverse sentence or passage level IR tasks. It provides a common and easy framework for the cross-domain evaluation of your retrieval models.

Recent Activity

BeIR's activity

nthakur 
posted an update 3 days ago
view post
Post
1469
Last year, I curated & generated a few multilingual SFT and DPO datasets by translating English SFT/DPO datasets into 9-10 languages using the mistralai/Mistral-7B-Instruct-v0.2 model.

I hope it helps the community for pretraining/instruction tuning multilingual LLMs! I added a small diagram to briefly describe which datasets are added and their sources.

Happy to collaborate in either using these datasets for instruction FT, or wishes to extend translated versions of newer SFT/DPO english datasets!

nthakur/multilingual-sft-and-dpo-datasets-67eaf56fe3feca5a57cf7d74
nthakur 
posted an update 11 months ago
view post
Post
3443
🦢 The SWIM-IR dataset contains 29 million text-retrieval training pairs across 27 diverse languages. It is one of the largest synthetic multilingual datasets generated using PaLM 2 on Wikipedia! 🔥🔥

SWIM-IR dataset contains three subsets :
- Cross-lingual:nthakur/swim-ir-cross-lingual
- Monolingual: nthakur/swim-ir-monolingual
- Indic Cross-lingual: nthakur/indic-swim-ir-cross-lingual

Check it out:
https://huggingface.co/collections/nthakur/swim-ir-dataset-662ddaecfc20896bf14dd9b7