Update README.md
Browse files
README.md
CHANGED
@@ -17,8 +17,9 @@ task_categories:
|
|
17 |
# ✨ ArcherCodeR
|
18 |
|
19 |
<div>
|
20 |
-
🏹️ Reinforcement Learning for
|
21 |
</div>
|
|
|
22 |
</div>
|
23 |
<div>
|
24 |
<br>
|
@@ -29,13 +30,12 @@ task_categories:
|
|
29 |
[](https://huggingface.co/wizardII/ArcherCodeR-1.5B)
|
30 |
[](https://huggingface.co/datasets/wizardII/ArcherCodeR-Dataset)
|
31 |
[](https://wandb.ai/wangjkpkucs-peking-university/ArcherCodeR?nw=nwuserwangjkpkucs)
|
32 |
-
|
33 |
-
</div>
|
34 |
|
35 |
</div>
|
36 |
|
37 |
|
38 |
-
##
|
39 |
|
40 |
[`ArcherCodeR-Dataset`](https://huggingface.co/datasets/wizardII/ArcherCodeR-Dataset) is **a dataset of verifiable, challenging, and diverse coding questions (6.7K)**. This dataset is used to train the **`ArcherCodeR`** model series, which consists of code reasoning models trained using large-scale rule-based reinforcement learning with carefully designed datasets and training recipes.
|
41 |
|
@@ -45,20 +45,13 @@ We select, clean, and curate coding problems from open-source datasets, includin
|
|
45 |
- [deepmind/code_contests](https://huggingface.co/datasets/deepmind/code_contests)
|
46 |
- [open-r1/codeforces](https://huggingface.co/datasets/open-r1/codeforces)
|
47 |
|
|
|
48 |
|
49 |
-
|
|
|
50 |
|
51 |
-
The technical report will be released soon.
|
52 |
|
53 |
-
##
|
54 |
|
55 |
-
|
56 |
|
57 |
-
```bibtex
|
58 |
-
@misc{archercoder2025,
|
59 |
-
title={ArcherCodeR},
|
60 |
-
author={Jiakang Wang},
|
61 |
-
note={Blog},
|
62 |
-
year={2025}
|
63 |
-
}
|
64 |
-
```
|
|
|
17 |
# ✨ ArcherCodeR
|
18 |
|
19 |
<div>
|
20 |
+
🏹️ Reinforcement Learning for Enhanced Code Reasoning in LLMs 🎯
|
21 |
</div>
|
22 |
+
|
23 |
</div>
|
24 |
<div>
|
25 |
<br>
|
|
|
30 |
[](https://huggingface.co/wizardII/ArcherCodeR-1.5B)
|
31 |
[](https://huggingface.co/datasets/wizardII/ArcherCodeR-Dataset)
|
32 |
[](https://wandb.ai/wangjkpkucs-peking-university/ArcherCodeR?nw=nwuserwangjkpkucs)
|
33 |
+
[](https://zhuanlan.zhihu.com/p/xxx)
|
|
|
34 |
|
35 |
</div>
|
36 |
|
37 |
|
38 |
+
## Overview
|
39 |
|
40 |
[`ArcherCodeR-Dataset`](https://huggingface.co/datasets/wizardII/ArcherCodeR-Dataset) is **a dataset of verifiable, challenging, and diverse coding questions (6.7K)**. This dataset is used to train the **`ArcherCodeR`** model series, which consists of code reasoning models trained using large-scale rule-based reinforcement learning with carefully designed datasets and training recipes.
|
41 |
|
|
|
45 |
- [deepmind/code_contests](https://huggingface.co/datasets/deepmind/code_contests)
|
46 |
- [open-r1/codeforces](https://huggingface.co/datasets/open-r1/codeforces)
|
47 |
|
48 |
+
### 🔍 Key Notes:
|
49 |
|
50 |
+
- Both code_contests (DeepMind) and codeforces (Open-r1) datasets include regenerated test cases to mitigate false positives.
|
51 |
+
- Significant prompt duplication exists across sources. When duplicates occur, code_contests or codeforces data takes priority.
|
52 |
|
|
|
53 |
|
54 |
+
## Technical Report
|
55 |
|
56 |
+
The technical report will be released soon.
|
57 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|