EPISODE · Mar 19, 2026 · 25 MIN
Online Experiential Learning for Language Models
from Daily Paper Cast · host Jingwen Liang, Gengyu Wang
🤗 Upvotes: 39 | cs.CL Authors: Tianzhu Ye, Li Dong, Qingxiu Dong, Xun Wu, Shaohan Huang, Furu Wei Title: Online Experiential Learning for Language Models Arxiv: http://arxiv.org/abs/2603.16856v1 Abstract: The prevailing paradigm for improving large language models relies on offline training with human annotations or simulated environments, leaving the rich experience accumulated during real-world deployment entirely unexploited. We propose Online Experiential Learning (OEL), a framework that enables language models to continuously improve from their own deployment experience. OEL operates in two stages: first, transferable experiential knowledge is extracted and accumulated from interaction trajectories collected on the user side; second, this knowledge is consolidated into model parameters via on-policy context distillation, requiring no access to the user-side environment. The two stages are iterated to form an online learning loop, where the improved model collects higher-quality trajectories that yield richer experiential knowledge for subsequent rounds. We evaluate OEL on text-based game environments across multiple model scales and both thinking and non-thinking variants. OEL achieves consistent improvements over successive iterations, enhancing both task accuracy and token efficiency while preserving out-of-distribution performance. Our analysis further shows that extracted experiential knowledge is significantly more effective than raw trajectories, and that on-policy consistency between the knowledge source and the policy model is critical for effective learning.
What this episode covers
🤗 Upvotes: 39 | cs.CL Authors: Tianzhu Ye, Li Dong, Qingxiu Dong, Xun Wu, Shaohan Huang, Furu Wei Title: Online Experiential Learning for Language Models Arxiv: http://arxiv.org/abs/2603.16856v1 Abstract: The prevailing paradigm for improving large language models relies on offline training with human annotations or simulated environments, leaving the rich experience accumulated during real-world deployment entirely unexploited. We propose Online Experiential Learning (OEL), a framework that enables language models to continuously improve from their own deployment experience. OEL operates in two stages: first, transferable experiential knowledge is extracted and accumulated from interaction trajectories collected on the user side; second, this knowledge is consolidated into model parameters via on-policy context distillation, requiring no access to the user-side environment. The two stages are iterated to form an online learning loop, where the improved model collects higher-quality trajectories that yield richer experiential knowledge for subsequent rounds. We evaluate OEL on text-based game environments across multiple model scales and both thinking and non-thinking variants. OEL achieves consistent improvements over successive iterations, enhancing both task accuracy and token efficiency while preserving out-of-distribution performance. Our analysis further shows that extracted experiential knowledge is significantly more effective than raw trajectories, and that on-policy consistency between the knowledge source and the policy model is critical for effective learning.
NOW PLAYING
Online Experiential Learning for Language Models
No transcript for this episode yet
Similar Episodes
No similar episodes found.