SDPO Trains Models With Natural-Language Feedback Instead of Scalar Rewards

• LLM — A large language model — the neural network behind tools like Claude and ChatGPT, trained on huge amounts of text to predict what comes next.

Offers an RL training approach that can use rich textual feedback (like compiler errors) directly as a training signal instead of collapsing it into a scalar reward.

• articleAI benchmarks aren’t as reliable as we think!
• postTTT-E2E Lets LLMs Keep Training on Context During Deployment
• postTheory of Space Benchmark Tests Active Spatial Exploration in AI
• articleCheck out our latest SAIL blog post on VAGEN, a reinforcement learning…

• articleGSPO: Towards Scalable Reinforcement Learning for Language ModelsQwen Team
• articleDirect Preference Optimization: Your Language Model is Secretly a Reward ModelRafael Rafailov et al.
• articleSERPO: Self-Evolving Rubric Policy Optimization for Open-Ended Test-Time Reinforcement LearningJianze Wang, Kunwang Zheng, Ying Liu, Yu Cao, Qilong Zhang, Jinlong Chen, Hua Yang, Qianglong Chen

Checking sign-in…

Loading comments…