devarsheegaunekar/

Konkani-Instruct-v2

A high-quality instruction-tuning dataset for building and fine-tuning Konkani Large Language Models (LLMs). This dataset is designed to improve:

Instruction following Conversational abilities Native Konkani understanding Cultural/domain knowledge General reasoning and response quality

Total Entries: 50,246

	Composition




	

Layer Description Entries

Layer 1 Alpaca-style instruction data 8,629

Layer 2 Native… See the full description on the dataset page: https://huggingface.co/datasets/devarsheegaunekar/Konkani-Instruct-v2.