Retrospective Learning from Interactions

Zizhao Chen, Mustafa Omer Gul, Yiwei Chen +3

Multi-turn interactions between large language models (LLMs) and users naturally include implicit feedback signals. If an LLM responds in an unexpected way to an instruction, the u…