Create README.md
Browse files
README.md
ADDED
|
@@ -0,0 +1,36 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
license: apache-2.0
|
| 3 |
+
datasets:
|
| 4 |
+
- Multiverse4FM/Autoregressive-1K-mixed
|
| 5 |
+
- Multiverse4FM/Multiverse-1K
|
| 6 |
+
- simplescaling/s1K-1.1
|
| 7 |
+
base_model:
|
| 8 |
+
- Qwen/Qwen2.5-32B-Instruct
|
| 9 |
+
pipeline_tag: text-generation
|
| 10 |
+
library_name: transformers
|
| 11 |
+
---
|
| 12 |
+
|
| 13 |
+
# Model Summary
|
| 14 |
+
|
| 15 |
+
> Autoregressive-32B is a baseline of our Multiverse-32B built on autoregressive modeling.
|
| 16 |
+
|
| 17 |
+
- **Webpage:** [Multiverse](https://multiverse4fm.github.io/)
|
| 18 |
+
- **Paper:** [https://arxiv.org/abs/2506.09991](https://arxiv.org/abs/2506.09991)
|
| 19 |
+
# Use
|
| 20 |
+
|
| 21 |
+
The model usage is documented [here](https://github.com/Multiverse4FM/Multiverse).
|
| 22 |
+
|
| 23 |
+
# Evaluation
|
| 24 |
+
|
| 25 |
+
| Model | AIME24 | AIME25 | MATH500 | GPQA-Diamond |
|
| 26 |
+
| :--- | :---: | :---: | :---: | :---: |
|
| 27 |
+
| s1-32B | 35.4 | 25.8 | 88.6 | 48.0 |
|
| 28 |
+
| s1.1-32B | 52.9 | 41.7 | 93.4 | 62.6 |
|
| 29 |
+
| Qwen2.5-32B-Instruct | 15.8 | 10.4 | 80.4 | 47.0 |
|
| 30 |
+
| Autoregressive-32B | **54.6** | <u>45.0</u> | **92.8** | <u>61.6</u> |
|
| 31 |
+
| **Multiverse-32B-zero** | 52.1 | 44.2 | <u>92.4</u> | **63.6** |
|
| 32 |
+
| **Multiverse-32B** | 53.8 | **45.8** | 91.8 | 60.7 |
|
| 33 |
+
|
| 34 |
+
# Acknowledge
|
| 35 |
+
|
| 36 |
+
Thanks to the amazing s1 team for their s1.1 dataset as base data, and the Qwen team for their Qwen-2.5-32B-Instruct as base model.
|