TIGER-Lab/AceCoder-Qwen2.5-Coder-7B-Ins-V1.1
TIGER-Lab/AceCoder-Qwen2.5-Coder-7B-Ins-V1.1 is a 7.6 billion parameter instruction-tuned causal language model developed by TIGER-Lab, based on the Qwen2.5-Coder-7B architecture. This model is specifically optimized for code generation and problem-solving, having been fine-tuned with Reinforcement Learning (RL) using the AceCode-V1.1-69K dataset. It demonstrates strong performance in coding benchmarks, even surpassing the base Qwen2.5-Coder-7B-Instruct model, making it suitable for advanced coding agent applications.
Loading preview...
AceCoder-Qwen2.5-Coder-7B-Ins-V1.1 Overview
AceCoder-Qwen2.5-Coder-7B-Ins-V1.1 is an advanced 7.6 billion parameter code generation model developed by TIGER-Lab. It is an updated version of the original AceCoder-Qwen2.5-Coder-7B-Base-Rule, built upon the Qwen Coder 7B Base model.
Key Capabilities & Training
This model's primary strength lies in its code generation and problem-solving abilities, achieved through a unique training methodology:
- Reinforcement Learning (RL) Fine-tuning: The model was fine-tuned using RL, specifically leveraging the binary pass rate as a rule-based reward signal.
- Specialized Dataset: Training was conducted on the proprietary TIGER-Lab/AceCode-V1.1-69K dataset, which is crucial for its enhanced coding performance.
Performance Highlights
AceCoder-V1.1-7B demonstrates competitive performance across various coding benchmarks, notably:
- LiveCodeBench-v4: Achieved 35.7, outperforming Qwen2.5-Coder-7B-Instruct (34.2).
- HumanEval: Scored 88.4, closely matching or exceeding other 7B models.
- MBPP: Recorded 84.9.
- BigCodeBench-Complete Full: Achieved 53.9, indicating strong performance on complex coding tasks.
These results highlight the effectiveness of TIGER-Lab's dataset and RL approach for developing highly capable coding agents.
Good For
- Code Generation: Excels at generating functional code snippets and solutions.
- Coding Agents: Ideal for integration into automated coding systems or development tools.
- Problem Solving: Particularly strong in tasks requiring logical reasoning and algorithmic implementation.