Dataset · KaLM-Embedding
Multilingual triplets for KaLM embedding training
KaLM-Embedding/KaLM-embedding-finetuning-data
From 1 to 10M examples with positive and negative pairs in English, Chinese, and other languages for embedding model fine-tuning.
What's inside
A feature-extraction dataset ranging from 1 to 10 million records in English, Chinese, and other languages. Each example is preprocessed and contains a query string, a list of relevant responses (pos, usually one), and a list of irrelevant examples (neg, usually seven).
What you can do with it
- Contrastive fine-tuning of text embedding models
- Improving semantic search quality and retrievers for RAG systems
- Tasks
- Embeddings
- Languages
- English, Chinese
- Size
- 1M–10M rows
- Downloads this month
- 524
- Likes
- 33
Models trained on it
- tencent/KaLM-Embedding-Gemma3-12B-2511Sentence similarity 9.2K
- mradermacher/KaLM-Embedding-Gemma3-12B-2511-i1-GGUF 2.5K
- KaLM-Embedding/KaLM-embedding-multilingual-mini-instruct-v2.5Embeddings 1.9K
- mradermacher/KaLM-Embedding-Gemma3-12B-2511-GGUF 758
- KaLM-Embedding/KaLM-Reranker-V1-Nano-R2Text ranking 560
- mradermacher/KaLM-embedding-multilingual-mini-instruct-v2.5-GGUF 507
- KaLM-Embedding/KaLM-Reranker-V1-Small-R2Text ranking 460
- KaLM-Embedding/KaLM-Reranker-V1-Large-R2Text ranking 440
How to load it
from datasets import load_dataset
ds = load_dataset("KaLM-Embedding/KaLM-embedding-finetuning-data")