Xiangyang Ji
4 indexed papers
Research Timeline
The paper introduces Group Prioritized Off-Policy Optimization (POPO), a novel framework that efficiently accelerates RL finetuning for LLM reasoning by leveraging effective off-policy training batches without requiring costly additional data rollouts.
The paper introduces Group Prioritized Off-Policy Optimization (POPO), a novel framework that efficiently accelerates RL finetuning for LLM reasoning by leveraging effective off-policy training batches without requiring costly additional data rollouts.
This paper introduces a new rollout allocation framework for enhancing reward contrast in reinforcement learning with large language models.
This paper presents Forge, a pipeline that uses formal verification to ensure the safety and certification of LLM-generated Java code.
Papers
Formal-Method-Guided Vibe Coding: Closing the Verification Loop on AI-Generated Safety-Critical Software Through Model-Driven Engineering
Ran Wei, Le Zhu, Haochi Wang, Jim Woodcock +3 more
This paper presents Forge, a pipeline that uses formal verification to ensure the safety and certification of LLM-generated Java code.