Update README.md

README.md

@@ -20,4 +20,61 @@ tags:
 - RAG
 license: apache-2.0
 library_name: sentence-transformers
----
+---
+
+
+<h1 align="center">KaLM-Reranker-V1: Fast but Not Late Interaction for Compressed Document Reranking</h1>
+
+<p align="center">
+  <a href="https://huggingface.co/collections/KaLM-Embedding/lychee-kalm-reranker">
+    <img src="https://img.shields.io/badge/%F0%9F%A4%97_Collection-KaLM--Reranker-ffbd45.svg" alt="Hugging Face Collection">
+  </a>
+  <a href="https://arxiv.org/abs/2606.22807">
+    <img src="https://img.shields.io/badge/arXiv-2606.22807-b31b1b.svg" alt="Paper">
+  </a>
+  <a href="https://github.com/KaLM-Embedding">
+    <img src="https://img.shields.io/badge/GitHub-KaLM--Embedding-blue.svg?logo=github" alt="GitHub">
+  </a>
+  <a href="https://www.apache.org/licenses/LICENSE-2.0">
+    <img src="https://img.shields.io/badge/License-Apache--2.0-green.svg" alt="License">
+  </a>
+</p>
+
+
+## What is R2?
+
+**KaLM-Reranker-V1-R2 is a substantially improved checkpoint release of KaLM-Reranker-V1.** It keeps the same fast-but-not-late-interaction (FBNL) architecture and inference interface as the original release, while introducing a stronger multi-stage training recipe for three practical goals:
+
+1. **Stronger compression robustness.** R2 supports Matryoshka embedding pooling (MEP) from **1x to 128x**, extending the maximum validated compression ratio from 32x to 128x. Even at 128x compression, all three model sizes retain at least 92% of their average nDCG@10 at 2x compression on both BEIR and MIRACL.
+2. **Adjustable test-time compute.** A single R2 checkpoint can trade compute for quality by changing the compression ratio. It can also use a coarse-to-fine cascade: cheaply screen all candidates with heavily compressed document representations, then spend more compute only on the most promising candidates.
+3. **Better multi-domain and multilingual ranking.** At the default 4x compression ratio, every R2 model improves on BEIR, while MIRACL average nDCG@10 increases by **+9.03**, **+7.17**, and **+4.85** points for Nano, Small, and Large, respectively.
+
+
+
+### R2 vs. the original release
+
+All results below use the same default compression ratio, `r=4`. Original-release results are taken from the previous model cards; R2 results are from the updated paper.
+
+| Model | BEIR original | BEIR R2 | Delta | MIRACL original | MIRACL R2 | Delta |
+|:--|--:|--:|--:|--:|--:|--:|
+| Nano | 57.41 | **58.54** | **+1.13** | 62.08 | **71.11** | **+9.03** |
+| Small | 60.01 | **61.07** | **+1.06** | 66.89 | **74.06** | **+7.17** |
+| Large | 62.87 | **63.53** | **+0.66** | 70.07 | **74.92** | **+4.85** |
+
+| Capability | Original release | R2 |
+|:--|:--|:--|
+| Validated MEP range | 1x-32x | **1x-128x** |
+| Test-time compute | Flexible MEP, not explicitly evaluated as a compute-scaling strategy | **Adjustable compression and evaluated coarse-to-fine cascading** |
+| Training | Original supervised checkpoints | **High-quality SFT -> soft-label distillation -> model soup** |
+| Multi-domain ranking | Strong | **Improved on BEIR for all three sizes** |
+| Multilingual ranking | Limited | **Substantially improved across MIRACL's 18 languages** |
+
+## Model family
+
+| Model | Activated parameters | Layers (encoder + decoder) | Sequence length | Document token dimension | MEP support | Instruction aware |
+|:--|--:|--:|--:|--:|:--:|:--:|
+| [KaLM-Reranker-V1-Nano-R2](https://huggingface.co/KaLM-Embedding/KaLM-Reranker-V1-Nano-R2) | 0.27B | 18 + 18 | 128K | 640 | 1x-128x | Yes |
+| [KaLM-Reranker-V1-Small-R2](https://huggingface.co/KaLM-Embedding/KaLM-Reranker-V1-Small-R2) | 1B | 26 + 26 | 128K | 1152 | 1x-128x | Yes |
+| [KaLM-Reranker-V1-Large-R2](https://huggingface.co/KaLM-Embedding/KaLM-Reranker-V1-Large-R2) | 4B | 34 + 34 | 128K | 2560 | 1x-128x | Yes |
+
+The reported sizes are **activated parameters**. Nano, Small, and Large are initialized from the T5Gemma2 270M-270M, 1B-1B, and 4B-4B encoder-decoder families, respectively.