Create README.md
Browse files
README.md
ADDED
@@ -0,0 +1,56 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
---
|
2 |
+
license: cc-by-nc-4.0
|
3 |
+
language:
|
4 |
+
- en
|
5 |
+
base_model:
|
6 |
+
- Qwen/Qwen3-4B
|
7 |
+
pipeline_tag: text-ranking
|
8 |
+
tags:
|
9 |
+
- finance
|
10 |
+
- legal
|
11 |
+
- code
|
12 |
+
- stem
|
13 |
+
- medical
|
14 |
+
---
|
15 |
+
# zerank-1: ZeroEntropy Inc.'s SoTA reranker
|
16 |
+
|
17 |
+
<!-- Provide a quick summary of what the model is/does. -->
|
18 |
+
|
19 |
+
This model is an open-weights reranker model meant to be integrated into RAG applications to rerank results from preliminary search methods such as embeddings, BM25, and hybrid search.
|
20 |
+
|
21 |
+
This reranker outperforms all competitor rerankers, including ones twice its size, across a wide variety of task domains, including on finance, legal, code, STEM, medical, and conversational data. See [this post](https://evals_blog_post) for more details.
|
22 |
+
This model is trained on an innovative multi-stage pipeline that models query-document relevance scores using adjusted Elo-like ratings. See [this post](https://technical_blog_post) and our Technical Report (Coming soon!) for more details.
|
23 |
+
|
24 |
+
|
25 |
+
|
26 |
+
|
27 |
+
## How to Use
|
28 |
+
|
29 |
+
## Evaluations
|
30 |
+
|
31 |
+
Comparing NDCG@10 starting from top 100 documents by embedding (using text-3-embedding-small):
|
32 |
+
|
33 |
+
| Task | Embedding | cohere-rerank-v3.5 | Salesforce/Llama-rank-v1 | zerank-1-small | zerank-1 |
|
34 |
+
|----------------|-----------|--------------------|--------------------------|----------------|----------|
|
35 |
+
| Code | 0.678 | 0.724 | 0.694 | 0.730 | 0.754 |
|
36 |
+
| Conversational | 0.250 | 0.571 | 0.484 | 0.556 | 0.596 |
|
37 |
+
| Finance | 0.839 | 0.824 | 0.828 | 0.861 | 0.894 |
|
38 |
+
| Legal | 0.703 | 0.804 | 0.767 | 0.817 | 0.821 |
|
39 |
+
| Medical | 0.619 | 0.750 | 0.719 | 0.773 | 0.796 |
|
40 |
+
| STEM | 0.401 | 0.510 | 0.595 | 0.680 | 0.694 |
|
41 |
+
|
42 |
+
Comparing BM25 and Hybrid Search without and with zerank-1:
|
43 |
+
|
44 |
+
<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"/>
|
45 |
+
|
46 |
+
|
47 |
+
## Citation
|
48 |
+
|
49 |
+
**BibTeX:**
|
50 |
+
|
51 |
+
Coming soon!
|
52 |
+
|
53 |
+
**APA:**
|
54 |
+
|
55 |
+
Coming soon!
|
56 |
+
|