Ring-Zero: Scaling reinforcement learning with verifiable rewards to 1 trillion parameters for emergent reasoning capabilities
Reinforcement learning with verifiable rewards without human-annotated data, often referred to as zero RL, has emerged as a powerful paradigm for eliciting chain-of-thought reasoning. However, due to…
Read the full articleYou might also wanna read
LLM-as-a-Verifier: A probabilistic verification framework for scaling LLM correctness assessment
Scaling pre-training, post-training, and test-time compute have become the central paradigms for improving the capabilities of LLMs. In this
arxiv.org·10d agoLLM-as-a-Verifier: A probabilistic verification framework for scaling LLM correctness assessment
Scaling pre-training, post-training, and test-time compute have become the central paradigms for improving the capabilities of LLMs. In this
arxiv.org·10d ago
QUBRIC: A Framework for Co-Designing Queries and Rubrics to Extend Reinforcement Learning Beyond Verifiable Rewards
Rubric-based RL is a promising route for extending reinforcement learning beyond verifiable rewards, yet existing methods optimize rubrics w
arxiv.org·1mo ago
Turing-RL: A Reinforcement Learning Approach for Training User Simulators Using Turing Test Rewards
Learning to simulate human users in interactive settings could advance the training of agent assistants, evaluation of personalization syste
arxiv.org·28d ago
SPIRAL: A Reinforcement Learning Framework for Multi-Primitive Language Model Reasoning
Language model reasoning can be substantially improved at test time via scaffolds that scale inference compute across different primitives -
arxiv.org·21d ago
Reinforcement Learning: A Double-Edged Sword in Continual AI Training
Reinforcement learning faces challenges in avoiding catastrophic forgetting in continual post-training for vision-language models. The new M
Reward Design Is the Hard Part: Building Verifiable Rewards for Tool-Using Agents
Photo by Todd Quackenbush on Unsplash RLVR works beautifully for math. Generate a response, extract the final answer, compare it to ground t
Comments
Sign in to join the conversation.
No comments yet. Be the first.