IrvinTopi/WebFAQHardNegatives
WebFAQ 2.0: Multilingual Hard Negatives This dataset contains mined hard negatives derived from the WebFAQ 2.0 corpus. It includes approximately 1.3 million samples across 20 languages. The dataset is designed to support robust training of dense retrieval models, specifically enabling: Contrastive Learning: Using strict hard negatives to improve discrimination. Knowledge Distillation: Using the provided cross-encoder scores to train with soft labels (e.g., MarginMSE).… See the full description on the dataset page: https://huggingface.co/datasets/IrvinTopi/WebFAQHardNegatives.
Upload mining_hardnegatives_bge3.py
Update README.md
Update README.md
Upload README.md with huggingface_hub
Upload data/zho/train_marginmse.jsonl with huggingface_hub
Upload data/vie/train_marginmse.jsonl with huggingface_hub
Upload data/tur/train_marginmse.jsonl with huggingface_hub
Upload data/swe/train_marginmse.jsonl with huggingface_hub
Upload data/spa/train_marginmse.jsonl with huggingface_hub
Upload data/rus/train_marginmse.jsonl with huggingface_hub
Upload data/por/train_marginmse.jsonl with huggingface_hub
Upload data/pol/train_marginmse.jsonl with huggingface_hub
Upload data/nld/train_marginmse.jsonl with huggingface_hub
Upload data/kor/train_marginmse.jsonl with huggingface_hub
Upload data/jpn/train_marginmse.jsonl with huggingface_hub
Upload data/ita/train_marginmse.jsonl with huggingface_hub
Upload data/ind/train_marginmse.jsonl with huggingface_hub
Upload data/hin/train_marginmse.jsonl with huggingface_hub
Upload data/fra/train_marginmse.jsonl with huggingface_hub
Upload data/fas/train_marginmse.jsonl with huggingface_hub
Upload data/eng/train_marginmse.jsonl with huggingface_hub
Upload data/deu/train_marginmse.jsonl with huggingface_hub
Upload data/dan/train_marginmse.jsonl with huggingface_hub
Upload data/ara/train_marginmse.jsonl with huggingface_hub
initial commit