README.md

---
title: KaLM Embedding
emoji: 🔎
colorFrom: pink
colorTo: indigo
sdk: gradio
sdk_version: 5.49.1
python_version: '3.12'
app_file: app.py
pinned: false
short_description: KaLM Embedding — Versatile Text Embedding
---

# KaLM Embedding —Versatile Text Embedding

An interactive demonstration of [KaLM-embedding-multilingual-mini-instruct-v2.5](https://huggingface.co/KaLM-Embedding/KaLM-embedding-multilingual-mini-instruct-v2.5).

KaLM-Embedding-V2 is a versatile and compact embedding model that achieves impressive performance in general-purpose text embedding tasks through systematic incentivization of advanced embedding techniques.

## What you can do

- **Semantic Retrieval:** paste a query and one document per line to rank the most relevant passages.
- **Semantic Textual Similarity:** compare the meaning of two pieces of text with the instruction `Retrieve semantically similar text.`
- **Embedding inspector:** create normalized vectors and view a compact preview of their values.
- **Task instructions:** optionally specify what a relevant result means for your retrieval task.

The Space uses 896-dimensional L2-normalized embeddings and cosine similarity. The model loads on first use; please keep public-demo inputs concise.

## Citation

```bibtex
@inproceedings{
  zhao2026kalmembeddingv,
  title={Ka{LM}-Embedding-V2: Superior Training Techniques and Data Inspire A Versatile Embedding Model},
  author={Xinping Zhao and Xinshuo Hu and Zifei Shan and Shouzheng Huang and Yao Zhou and Xin Zhang and Zetian Sun and Zhenyu Liu and Dongfang Li and Xinyuan Wei and Youcheng Pan and Yang Xiang and Meishan Zhang and Haofen Wang and Jun Yu and Baotian Hu and Min Zhang},
  booktitle={The Fourteenth International Conference on Learning Representations},
  year={2026},
  url={https://openreview.net/forum?id=Y7qzhvWhcz}
}
```