Hugging Face's logo

THU-KEG
/
LLaDA-8B-BGPO-sudoku

Reinforcement Learning
Safetensors
English
llada
sudoku
dllm
bgpo
custom_code
Model card Files
xet
Community
1
LLaDA-8B-BGPO-sudoku
File size: 1,328 Bytes 
b691968
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
101ce15
 
b691968
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
 
---
license: apache-2.0
language:
- en
tags:
- reinforcement-learning
- sudoku
- dllm
- bgpo
- llada
size_categories:
- 8B
---

# LLaDA-8B-BGPO-sudoku

[![Paper](https://img.shields.io/badge/Paper-arXiv:2510.11683-red)](https://arxiv.org/abs/2510.11683)
[![Code](https://img.shields.io/badge/Code-GitHub-blue)](https://github.com/THU-KEG/BGPO)

## Model Description

**LLaDA-8B-BGPO-sudoku** is an 8-billion parameter diffusion large language model (dLLM) that was trained on LLaDA-8B-Instruct using Boundary-Guided Policy Optimization (BGPO) for enhanced Sudoku solving capabilities.

## Model Details

- **Model Type**: Diffusion Large Language Model (dLLM)
- **Parameters**: 8 billion
- **Training Method**: Boundary-Guided Policy Optimization (BGPO)
- **Base Model**: LLaDA-8B-Instruct
- **Task**: Sudoku
- **Language**: English

## Training Details

- **Training Steps**: 400 steps
- **Response Length**: 256 tokens
- **Train Diffusion Steps**: 128
- **Eval Diffusion Steps**: 256
- **Block Size**: 32
- **Monte Carlo Sample Size ($n_t$)**: 32
- **Learning Rate**: 5e-7
- **Batch Size**: 16
- **Framework**: Built on VeRL (Volcengine Reinforcement Learning)

## Usage & Limitations

- Primarily designed for Sudoku tasks.
- Performance may vary on other tasks.
- Requires appropriate computational resources for inference.