Spaces:
Running on Zero
Running on Zero
| title: KaLM Reranker | |
| emoji: 🚀 | |
| colorFrom: blue | |
| colorTo: indigo | |
| sdk: gradio | |
| sdk_version: 6.20.0 | |
| python_version: '3.12' | |
| app_file: app.py | |
| pinned: false | |
| license: apache-2.0 | |
| short_description: Fast but Not Late Interaction Reranker | |
| # KaLM Reranker — Compressed Document Reranking | |
| An interactive demo for [KaLM-Reranker-V1-Small](https://huggingface.co/KaLM-Embedding/KaLM-Reranker-V1-Small), a fast but not late-interaction reranker for efficient query-document relevance scoring. | |
| ## Features | |
| - **Semantic Reranking:** rank a set of documents for one query. | |
| - **Pair Relevance:** return the relevance probability for one query-document pair. | |
| - **Compression Explorer:** compare relevance scores and estimated encoder representation lengths at 1×–32× compression. | |
| The compression factor maps to the official `chunk_size` hyperparameter. Larger values pool more encoder token states into one chunk, reducing the document representation passed to the decoder. | |
| ## Citation | |
| ```bibtex | |
| @misc{zhao2026kalmrerankerv1, | |
| title={KaLM-Reranker-V1: Fast but Not Late Interaction for Compressed Document Reranking}, | |
| author={Xinping Zhao and Jiaxin Xu and Ziqi Dai and Xin Zhang and Shouzheng Huang and Danyu Tang and Xinshuo Hu and Meishan Zhang and Baotian Hu and Min Zhang}, | |
| year={2026}, | |
| eprint={2606.22807}, | |
| archivePrefix={arXiv}, | |
| primaryClass={cs.CL}, | |
| url={https://arxiv.org/abs/2606.22807} | |
| } | |
| ``` |