INSAIT-Institute/BatchOnline-GRPO-1.5B

TEXT GENERATIONPricing:Input $0.04 / Cached $0.008 / Output $0.08Concurrent Unit Cost:1Model Size:1.5BQuant:BF16Context Size:32kTool Calling:SupportedPublished:Aug 12, 2026License:cc-by-nc-sa-4.0Architecture:Transformer Open Weights Featherless Exclusive Cold

INSAIT-Institute/BatchOnline-GRPO-1.5B is a 1.5 billion parameter code verifier model developed by INSAIT-Institute, fine-tuned from DeepSeek-R1-Distill-Qwen-1.5B. This model is trained with semi-online GRPO, where the generator policy is synced every four steps, and is designed to judge and rank candidate solutions for competitive programming problems. It excels at identifying correct code snippets and is optimized for integration into post-training pipelines for large code generation models.

Loading preview...

Aletheia: BatchOnline-GRPO-1.5B Code Verifier

This model, BatchOnline-GRPO-1.5B, is a 1.5 billion parameter code verifier developed by INSAIT-Institute. It is fine-tuned from deepseek-ai/DeepSeek-R1-Distill-Qwen-1.5B and trained using a semi-online variant of Generative Reinforcement Policy Optimization (GRPO), where the generator policy is synchronized every four steps. This approach is part of the broader Aletheia project, which investigates efficient training recipes for robust code verifiers.

Key Capabilities

  • Code Verification: Judges and ranks candidate solutions for competitive programming problems, identifying the correct code snippet among several options.
  • Reinforcement Learning with Verifiable Rewards (RLVR): Utilizes an RLVR framework to train verifiers, focusing on the trade-offs between performance and cost.
  • Scalable Training Insights: Part of a research effort demonstrating that optimal training recipes for verifiers are scale-dependent, with on-policy learning being crucial for smaller models and thinking budget for larger ones.
  • Negative Sample Leveraging: Incorporates learning from negative samples, which consistently improves top-1 selection accuracy and stabilizes training at larger scales.

Good for

  • RLHF / RLAIF: Serves as a plug-and-play reward function for optimizing code generation policies.
  • Automated Evaluation: Can be used as an LLM-as-a-judge for various code-related tasks.
  • Research: Ideal for studying the impact of thinking traces, on-policy learning, and negative samples in developing effective code verifiers.

Finetunes
1 models