README.md

---
base_model:
- Qwen/Qwen3-4B
language:
- en
library_name: sentence-transformers
license: cc-by-nc-4.0
pipeline_tag: text-ranking
tags:
- finance
- legal
- code
- stem
- medical
---

<img src="https://i.imgur.com/oxvhvQu.png"/>

# Releasing zeroentropy/zerank-1

This model is the `zerank-1` reranker as introduced in the paper [zELO: ELO-inspired Training Method for Rerankers and Embedding Models](https://huggingface.co/papers/2509.12541).

**Code**: [https://github.com/zeroentropy-ai/zbench](https://github.com/zeroentropy-ai/zbench)

In search engines, [rerankers are crucial](https://www.zeroentropy.dev/blog/what-is-a-reranker-and-do_i_need_one) for improving the accuracy of your retrieval system.

However, SOTA rerankers are closed-source and proprietary. At ZeroEntropy, we've trained a SOTA reranker outperforming closed-source competitors, and we're launching our model here on HuggingFace.

This reranker [outperforms proprietary rerankers](https://huggingface.co/zeroentropy/zerank-1#evaluations) such as `cohere-rerank-v3.5` and `Salesforce/LlamaRank-v1` across a wide variety of domains, including finance, legal, code, STEM, medical, and conversational data.

At ZeroEntropy we've developed an innovative multi-stage pipeline that models query-document relevance scores as adjusted [Elo ratings](https://en.wikipedia.org/wiki/Elo_rating_system). More details are available in our paper: [zELO: ELO-inspired Training Method for Rerankers and Embedding Models](https://huggingface.co/papers/2509.12541).

Since we're a small company, this model is only released under a non-commercial license. If you'd like a commercial license, please contact us at founders@zeroentropy.dev and we'll get you a license ASAP.

For this model's smaller twin, see [zerank-1-small](https://huggingface.co/zeroentropy/zerank-1-small), which we've fully open-sourced under an Apache 2.0 License.

## How to Use

```python
from sentence_transformers import CrossEncoder

model = CrossEncoder("zeroentropy/zerank-1", trust_remote_code=True)

query_documents = [
    ("What is 2+2?", "4"),
    ("What is 2+2?", "The answer is definitely 1 million"),
]

scores = model.predict(query_documents)

print(scores)
```

The model can also be inferenced using ZeroEntropy's [/models/rerank](https://docs.zeroentropy.dev/api-reference/models/rerank) endpoint.

## Evaluations

NDCG@10 scores between `zerank-1` and competing closed-source proprietary rerankers. Since we are evaluating rerankers, OpenAI's `text-embedding-3-small` is used as an initial retriever for the Top 100 candidate documents.

| Task | Embedding | cohere-rerank-v3.5 | Salesforce/Llama-rank-v1 | zerank-1-small | **zerank-1** |
|---|---|---|---|---|---|
| Code | 0.678 | 0.724 | 0.694 | 0.730 | **0.754** |
| Conversational | 0.250 | 0.571 | 0.484 | 0.556 | **0.596** |
| Finance | 0.839 | 0.824 | 0.828 | 0.861 | **0.894** |
| Legal | 0.703 | 0.804 | 0.767 | 0.817 | **0.821** |
| Medical | 0.619 | 0.750 | 0.719 | 0.773 | **0.796** |
| STEM | 0.401 | 0.510 | 0.595 | 0.680 | **0.694** |

Comparing BM25 and Hybrid Search without and with zerank-1:

<img src="https://cdn-uploads.huggingface.co/production/uploads/67776f9dcd9c9435499eafc8/2GPVHFrI39FspnSNklhsM.png" alt="Description" width="400"/> <img src="https://cdn-uploads.huggingface.co/production/uploads/67776f9dcd9c9435499eafc8/dwYo2D7hoL8QiE8u3yqr9.png" alt="Description" width="400"/>