Improve model card for Re:Form: add metadata, links, and usage example
#1
by
nielsr
HF Staff
- opened
README.md
CHANGED
@@ -1,3 +1,83 @@
|
|
1 |
-
---
|
2 |
-
license: mit
|
3 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
---
|
2 |
+
license: mit
|
3 |
+
pipeline_tag: text-generation
|
4 |
+
library_name: transformers
|
5 |
+
tags:
|
6 |
+
- code-generation
|
7 |
+
- formal-verification
|
8 |
+
- reinforcement-learning
|
9 |
+
- dafny
|
10 |
+
---
|
11 |
+
|
12 |
+
# Re:Form -- Reducing Human Priors in Scalable Formal Software Verification with RL in LLMs: A Preliminary Study on Dafny
|
13 |
+
|
14 |
+
This repository hosts the `sft_0.5B` model, part of the **Re:Form** framework, which focuses on enhancing formal software verification using Large Language Models (LLMs) and Reinforcement Learning (RL).
|
15 |
+
|
16 |
+
**Re:Form** addresses the limitations of informal language-based LLMs by grounding them in rigorous formal systems, such as Dafny. This approach enables automatic and mathematically provable verification of the LLMs' reasoning processes and outcomes, crucial for large-scale, reliable formal software verification. The framework introduces an automatic and scalable data curation pipeline and integrates RL designs with feedback from a formal language verifier to reduce reliance on human priors.
|
17 |
+
|
18 |
+
- **Paper**: [Re:Form -- Reducing Human Priors in Scalable Formal Software Verification with RL in LLMs: A Preliminary Study on Dafny](https://huggingface.co/papers/2507.16331)
|
19 |
+
- **Project Page**: [https://veri-code.github.io/ReForm-page](https://veri-code.github.io/ReForm-page)
|
20 |
+
- **Code**: [https://github.com/Veri-Code/Veri-Code](https://github.com/Veri-Code/Veri-Code)
|
21 |
+
|
22 |
+
<p align="center">
|
23 |
+
<img src="https://github.com/Veri-Code/Veri-Code/raw/main/assets/pipeline.png" alt="Overall Pipeline" width="700"/>
|
24 |
+
<br>
|
25 |
+
<em>Overall Pipeline of Re:Form</em>
|
26 |
+
</p>
|
27 |
+
|
28 |
+
## Usage
|
29 |
+
|
30 |
+
You can load and use the `Re:Form` model with the `transformers` library for Dafny code generation.
|
31 |
+
|
32 |
+
```python
|
33 |
+
from transformers import AutoModelForCausalLM, AutoTokenizer
|
34 |
+
import torch
|
35 |
+
|
36 |
+
model_id = "Veri-Code/sft_0.5B" # Example checkpoint; other available include sft_1.5B, sft_3B, sft_7B, sft_14B, 14B-RL-entropy
|
37 |
+
|
38 |
+
tokenizer = AutoTokenizer.from_pretrained(model_id, trust_remote_code=True)
|
39 |
+
model = AutoModelForCausalLM.from_pretrained(
|
40 |
+
model_id,
|
41 |
+
torch_dtype=torch.bfloat16, # Use bfloat16 for optimal performance if supported
|
42 |
+
device_map="auto",
|
43 |
+
trust_remote_code=True,
|
44 |
+
)
|
45 |
+
|
46 |
+
# Example prompt for Dafny code generation
|
47 |
+
# This prompt asks the model to implement a simple Max method in Dafny.
|
48 |
+
prompt = "method Max(a: int, b: int) returns (m: int)\
|
49 |
+
ensures m == a || m == b\
|
50 |
+
ensures m >= a && m >= b\
|
51 |
+
{\
|
52 |
+
"
|
53 |
+
|
54 |
+
input_ids = tokenizer(prompt, return_tensors="pt").to(model.device)
|
55 |
+
|
56 |
+
# Generate Dafny code
|
57 |
+
generated_ids = model.generate(
|
58 |
+
**input_ids,
|
59 |
+
max_new_tokens=100,
|
60 |
+
temperature=0.7,
|
61 |
+
do_sample=True,
|
62 |
+
eos_token_id=tokenizer.eos_token_id,
|
63 |
+
)
|
64 |
+
|
65 |
+
generated_text = tokenizer.decode(generated_ids[0], skip_special_tokens=True)
|
66 |
+
print(generated_text)
|
67 |
+
```
|
68 |
+
|
69 |
+
## Citation
|
70 |
+
|
71 |
+
If you use this work in your research, please cite:
|
72 |
+
|
73 |
+
```bibtex
|
74 |
+
@misc{yan2025reformreducinghuman,
|
75 |
+
title={Re:Form -- Reducing Human Priors in Scalable Formal Software Verification with RL in LLMs: A Preliminary Study on Dafny},
|
76 |
+
author={Chuanhao Yan and Fengdi Che and Xuhan Huang and Xu Xu and Xin Li and Yizhi Li and Xingwei Qu and Jingzhe Shi and Zhuangzhuang He and Chenghua Lin and Yaodong Yang and Binhang Yuan and Hang Zhao and Yu Qiao and Bowen Zhou and Jie Fu},
|
77 |
+
year={2025},
|
78 |
+
eprint={2507.16331},
|
79 |
+
archivePrefix={arXiv},
|
80 |
+
primaryClass={cs.CL},
|
81 |
+
url={https://arxiv.org/abs/2507.16331},
|
82 |
+
}
|
83 |
+
```
|