Commit
•
99f6520
1
Parent(s):
59d99d9
Update README.md (#1)
Browse files- Update README.md (3d6574dad48c3add16d77c0423777e5e28ca3226)
Co-authored-by: Satya <[email protected]>
README.md
CHANGED
@@ -1,3 +1,45 @@
|
|
1 |
-
---
|
2 |
-
license: apache-2.0
|
3 |
-
---
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
---
|
2 |
+
license: apache-2.0
|
3 |
+
---
|
4 |
+
|
5 |
+
# FRAMES: Factuality, Retrieval, And reasoning MEasurement Set
|
6 |
+
|
7 |
+
FRAMES is a comprehensive evaluation dataset designed to test the capabilities of Retrieval-Augmented Generation (RAG) systems across factuality, retrieval accuracy, and reasoning.
|
8 |
+
|
9 |
+
## Dataset Overview
|
10 |
+
|
11 |
+
- 824 challenging multi-hop questions requiring information from 2-15 Wikipedia articles
|
12 |
+
- Questions span diverse topics including history, sports, science, animals, health, etc.
|
13 |
+
- Each question is labeled with reasoning types: numerical, tabular, multiple constraints, temporal, and post-processing
|
14 |
+
- Gold answers and relevant Wikipedia articles provided for each question
|
15 |
+
|
16 |
+
## Key Features
|
17 |
+
|
18 |
+
- Tests end-to-end RAG capabilities in a unified framework
|
19 |
+
- Requires integration of information from multiple sources
|
20 |
+
- Incorporates complex reasoning and temporal disambiguation
|
21 |
+
- Designed to be challenging for state-of-the-art language models
|
22 |
+
|
23 |
+
## Usage
|
24 |
+
|
25 |
+
This dataset can be used to:
|
26 |
+
- Evaluate RAG system performance
|
27 |
+
- Benchmark language model factuality and reasoning
|
28 |
+
- Develop and test multi-hop retrieval strategies
|
29 |
+
|
30 |
+
## Baseline Results
|
31 |
+
|
32 |
+
We provide baseline results using state-of-the-art models like Gemini-Pro-1.5-0514:
|
33 |
+
|
34 |
+
- Naive prompting: 40.8% accuracy
|
35 |
+
- BM25 retrieval (4 docs): 47.4% accuracy
|
36 |
+
- Oracle retrieval: 72.9% accuracy
|
37 |
+
- Multi-step retrieval & reasoning: 66% accuracy
|
38 |
+
|
39 |
+
## Citation
|
40 |
+
|
41 |
+
If you use this dataset in your research, please cite our paper:
|
42 |
+
|
43 |
+
|
44 |
+
|
45 |
+
We hope FRAMES will be useful for advancing RAG systems and language model capabilities. For more details, please refer to our full paper.
|