Add license, library_name and pipeline_tag
Browse filesThis PR adds the `library_name`, `license` and `pipeline_tag` metadata, and ensures the model can be found when filtering the hub for models compatible with the Transformers library. The technical report is also now released and the model card is updated accordingly.
README.md
CHANGED
|
@@ -1,8 +1,11 @@
|
|
| 1 |
---
|
| 2 |
-
datasets:
|
| 3 |
-
- Skywork/Skywork-OR1-RL-Data
|
| 4 |
base_model:
|
| 5 |
- deepseek-ai/DeepSeek-R1-Distill-Qwen-32B
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 6 |
---
|
| 7 |
|
| 8 |
# 🤔 Skywork-OR1 (Open Reasoner 1)
|
|
@@ -30,7 +33,7 @@ base_model:
|
|
| 30 |
- 🤗 Training data: [`Skywork-OR1-RL-Data`](https://huggingface.co/datasets/Skywork/Skywork-OR1-RL-Data)
|
| 31 |
- 🧑💻 Code: [`Skywork-OR1`](https://github.com/SkyworkAI/Skywork-OR1)
|
| 32 |
|
| 33 |
-
The complete technical report is
|
| 34 |
|
| 35 |
## 📖 Overview
|
| 36 |
|
|
@@ -53,6 +56,7 @@ The **`Skywork-OR1`** (Open Reasoner 1) model series consists of powerful math a
|
|
| 53 |
<img src="./assets/figure_2.jpeg" width="75%"/>
|
| 54 |
</div>
|
| 55 |
</div>
|
|
|
|
| 56 |
|
| 57 |
We evaluate our models on AIME24, AIME25, and LiveCodeBench. Instead of using Pass@1, which is common in prior work, we introduce Avg@K as the primary metric. This metric robustly measures a model's average performance across K independent attempts, reducing the impact of randomness and enhancing the reliability of the results. We believe that Avg@K provides a better reflection of a model's stability and reasoning consistency.
|
| 58 |
|
|
@@ -91,7 +95,7 @@ We develop a customized version of GRPO that leverages both data-wise and traini
|
|
| 91 |
|
| 92 |
## 📄 Technical Report
|
| 93 |
|
| 94 |
-
Our technical report
|
| 95 |
|
| 96 |
## 🙏 Acknowledgements
|
| 97 |
|
|
|
|
| 1 |
---
|
|
|
|
|
|
|
| 2 |
base_model:
|
| 3 |
- deepseek-ai/DeepSeek-R1-Distill-Qwen-32B
|
| 4 |
+
datasets:
|
| 5 |
+
- Skywork/Skywork-OR1-RL-Data
|
| 6 |
+
license: apache-2.0
|
| 7 |
+
library_name: transformers
|
| 8 |
+
pipeline_tag: text-generation
|
| 9 |
---
|
| 10 |
|
| 11 |
# 🤔 Skywork-OR1 (Open Reasoner 1)
|
|
|
|
| 33 |
- 🤗 Training data: [`Skywork-OR1-RL-Data`](https://huggingface.co/datasets/Skywork/Skywork-OR1-RL-Data)
|
| 34 |
- 🧑💻 Code: [`Skywork-OR1`](https://github.com/SkyworkAI/Skywork-OR1)
|
| 35 |
|
| 36 |
+
The complete technical report is now available: see our [Technical Report](https://huggingface.co/papers/2505.22312). See our [Notion Blog](https://capricious-hydrogen-41c.notion.site/Skywork-Open-Reaonser-Series-1d0bc9ae823a80459b46c149e4f51680) for training recipes and preliminary experimental results. More analysis and insights will be included in the final technical report, which is dedicated to helping the community better research, understand, and push the frontier of open reasoning models.
|
| 37 |
|
| 38 |
## 📖 Overview
|
| 39 |
|
|
|
|
| 56 |
<img src="./assets/figure_2.jpeg" width="75%"/>
|
| 57 |
</div>
|
| 58 |
</div>
|
| 59 |
+
<br>
|
| 60 |
|
| 61 |
We evaluate our models on AIME24, AIME25, and LiveCodeBench. Instead of using Pass@1, which is common in prior work, we introduce Avg@K as the primary metric. This metric robustly measures a model's average performance across K independent attempts, reducing the impact of randomness and enhancing the reliability of the results. We believe that Avg@K provides a better reflection of a model's stability and reasoning consistency.
|
| 62 |
|
|
|
|
| 95 |
|
| 96 |
## 📄 Technical Report
|
| 97 |
|
| 98 |
+
Our technical report is now available: see [Skywork Open Reasoner 1 Technical Report](https://huggingface.co/papers/2505.22312).
|
| 99 |
|
| 100 |
## 🙏 Acknowledgements
|
| 101 |
|