Emotional RAG LLMs: Reading Comprehension for the Open Internet
This addresses a challenge for real-world RAG applications that retrieve non-neutral internet text, but it is incremental as it builds on existing RAG frameworks.
The paper tackled the problem of retrieval-augmented generation (RAG) systems struggling with internet-based text in diverse emotional tones by introducing a dataset of emotionally inflected passages, an emotion translation model, and a prompt-based method, resulting in improved LLM interpretation of such text.
Queries to large language models (LLMs) can be divided into two parts: the instruction/question and the accompanying context. The context for retrieval-augmented generation (RAG) systems in most benchmarks comes from Wikipedia-like texts written in a neutral and factual tone. However, real-world RAG applications often retrieve internet-based text with diverse tones and linguistic styles, posing challenges for downstream tasks. This paper introduces (a) a dataset that transforms RAG-retrieved passages into emotionally inflected and sarcastic text, (b) an emotion translation model for adapting text to different tones, and (c) a prompt-based method to improve LLMs' pragmatic interpretation of retrieved text.