INSAIT-Institute/GRPO-Think-7B-8k
INSAIT-Institute/GRPO-Think-7B-8k is a 7.6 billion parameter GRPO-Think code verifier model developed by INSAIT-Institute. Fine-tuned from DeepSeek-R1-Distill-Qwen-7B, it is trained with Reinforcement Learning with Verifiable Rewards (RLVR) using on-policy learning, negative samples, and intermediate thinking traces, with an 8k reasoning-token budget. This model excels at judging and ranking candidate solutions for competitive programming problems, serving as a plug-and-play reward function for code generation policy optimization and automated evaluation.
Loading preview...
Overview
INSAIT-Institute/GRPO-Think-7B-8k is a 7.6 billion parameter code verifier model, part of the Aletheia project by INSAIT-Institute. It is fine-tuned from deepseek-ai/DeepSeek-R1-Distill-Qwen-7B and trained using GRPO (Generative Reinforcement Learning with Policy Optimization) with an 8k reasoning-token budget. This model is specifically designed to judge and rank candidate solutions for competitive programming problems, based on an execution-grounded testbed.
Key Capabilities
- Code Verification: Judges and ranks multiple candidate solutions for a given competitive programming problem.
- RLVR Training: Utilizes Reinforcement Learning with Verifiable Rewards (RLVR), incorporating on-policy learning, negative samples, and intermediate thinking traces for robust training.
- Scalability Insights: Developed as part of research exploring the performance-cost trade-offs in RLVR for code verifiers across different model sizes.
Good For
- RLHF / RLAIF: Serving as a 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 training effective code verifiers.