Aletheia-Bench/BatchOnline-GRPO-7B

TEXT GENERATIONPricing:Input $0.4 / Cached $0.08 / Output $0.8Concurrent Unit Cost:1Model Size:7.6BQuant:FP8Context Size:32kTool Calling:SupportedPublished:Dec 9, 2025License:cc-by-nc-sa-4.0Architecture:Transformer0.0K Open Weights Featherless Exclusive Cold

Aletheia-Bench/BatchOnline-GRPO-7B is a 7.6 billion parameter code verifier model developed by Aletheia-Bench, fine-tuned from DeepSeek-R1-Distill-Qwen-7B. This model is trained with semi-online GRPO, where the generator policy is synced every four steps, making it distinct in its training methodology for code verification. It excels at judging and ranking candidate solutions for competitive programming problems, providing a robust reward function for code generation policy optimization.

Loading preview...

Aletheia: BatchOnline-GRPO-7B Code Verifier

BatchOnline-GRPO-7B is a 7.6 billion parameter code verifier model developed by Aletheia-Bench, fine-tuned from deepseek-ai/DeepSeek-R1-Distill-Qwen-7B. This model is specifically designed for judging and ranking candidate solutions to competitive programming problems. Its unique characteristic lies in its training approach: it utilizes semi-online GRPO, where the generator policy is synchronized every four steps, offering a distinct balance between performance and computational cost in Reinforcement Learning with Verifiable Rewards (RLVR) pipelines.

Key Capabilities

  • Code Verification: Judges and ranks candidate code solutions for competitive programming problems.
  • RLVR Optimization: Provides a plug-and-play reward function for optimizing code generation policies through RLHF/RLAIF.
  • Automated Evaluation: Can function as an LLM-as-a-judge for various code-related tasks.
  • Research Testbed: Serves as a controlled environment for studying the impact of thinking traces, on-policy learning, and negative samples in training effective code verifiers.

Training & Datasets

The model was trained on the Aletheia-Bench/Aletheia-Train dataset, which comprises 50,000 training instances of competitive programming problems paired with candidate solutions. The Aletheia dataset collection also includes specialized datasets like Aletheia-DPO for preference optimization and Aletheia-Mixed for adversarial modifications, facilitating comprehensive verifier training and analysis.

Finetunes
1 models