CoGen: Learning from Feedback with Coupled Comprehension and Generation
This work addresses the challenge of making AI language systems more effective and human-like through interactive learning, though it is incremental as it builds on existing two-player game frameworks.
The paper tackled the problem of improving language systems by coupling comprehension and generation to learn from user feedback in reference games, resulting in performance improvements up to 26% in absolute terms and 17% higher accuracies compared to non-coupled systems.
Systems with both language comprehension and generation capabilities can benefit from the tight connection between the two. This work studies coupling comprehension and generation with focus on continually learning from interaction with users. We propose techniques to tightly integrate the two capabilities for both learning and inference. We situate our studies in two-player reference games, and deploy various models for thousands of interactions with human users, while learning from interaction feedback signals. We show dramatic improvements in performance over time, with comprehension-generation coupling leading to performance improvements up to 26% in absolute terms and up to 17% higher accuracies compared to a non-coupled system. Our analysis also shows coupling has substantial qualitative impact on the system's language, making it significantly more human-like.