Aletheia-Bench/GRPO-Think-7B-16k
Aletheia-Bench/GRPO-Think-7B-16k is a 7.6 billion parameter GRPO-Think verifier model developed by Aletheia-Bench, fine-tuned from DeepSeek-R1-Distill-Qwen-7B. It is trained with Reinforcement Learning with Verifiable Rewards (RLVR) using a 16k reasoning-token budget to judge and rank candidate code solutions for competitive programming problems. This model excels at providing a plug-and-play reward function for code generation policy optimization and automated evaluation as an LLM-as-a-judge.
Loading preview...
Aletheia-Bench/GRPO-Think-7B-16k Overview
GRPO-Think-7B-16k is a 7.6 billion parameter code verifier model developed by Aletheia-Bench. It is fine-tuned from deepseek-ai/DeepSeek-R1-Distill-Qwen-7B and trained using the GRPO (Generative Reinforcement Learning with Policy Optimization) method, which incorporates on-policy learning, negative samples, and intermediate thinking traces. A key feature of this model is its 16k reasoning-token budget (maximum completion length) used during training, enabling it to process and generate detailed thought processes when evaluating code.
Key Capabilities
- Code Verification: Judges and ranks candidate solutions for competitive programming problems, identifying the correct one from a set of options.
- Reinforcement Learning with Verifiable Rewards (RLVR): Utilizes a sophisticated training pipeline that includes intermediate thinking traces, learning from negative samples, and on-policy training to enhance verification accuracy.
- Multi-domain Thinking: Designed to perform multi-domain thinking, crucial for robust code verification.
Good For
- RLHF / RLAIF: Serving as a plug-and-play reward function for optimizing code generation policies.
- Automated Evaluation: Acting as an LLM-as-a-judge for various code-related tasks.
- Research: Studying the impact of thinking traces, on-policy learning, and negative samples in the development of effective code verifiers.