TY - GEN
T1 - CAF-I: A Collaborative Multi-Agent Framework for Enhanced Irony Detection with Large Language Models
AU - Liu, Ziqi
AU - Zhou, Ziyang
AU - Hu, Mingxuan
AU - Chen, Yangbin
AU - Xu, Zhijie
PY - 2025
Y1 - 2025
N2 - Large language model (LLM) have become mainstream methods in the field of sarcasm detection. However, existing LLM methods face challenges in irony detection, including: 1. single-perspective limitations, 2. insufficient comprehensive understanding, and 3. lack of interpretability. This paper introduces the Collaborative Agent Framework for Irony (CAF-I), an LLM-driven multi-agent system designed to overcome these issues. CAF-I employs specialized agents for Context, Semantics, and Rhetoric, which perform multidimensional analysis and engage in interactive collaborative optimization. A Decision Agent then consolidates these perspectives, with a Refinement Evaluator Agent providing conditional feedback for optimization. Experiments on benchmark datasets establish CAF-I's state-of-the-art zero-shot performance. Achieving SOTA on the vast majority of metrics, CAF-I reaches an average Macro-F1 of 76.31%, a 4.98% absolute improvement over the strongest prior baseline. This success is attained by its effective simulation of human-like multi-perspective analysis, enhancing detection accuracy and interpretability.
AB - Large language model (LLM) have become mainstream methods in the field of sarcasm detection. However, existing LLM methods face challenges in irony detection, including: 1. single-perspective limitations, 2. insufficient comprehensive understanding, and 3. lack of interpretability. This paper introduces the Collaborative Agent Framework for Irony (CAF-I), an LLM-driven multi-agent system designed to overcome these issues. CAF-I employs specialized agents for Context, Semantics, and Rhetoric, which perform multidimensional analysis and engage in interactive collaborative optimization. A Decision Agent then consolidates these perspectives, with a Refinement Evaluator Agent providing conditional feedback for optimization. Experiments on benchmark datasets establish CAF-I's state-of-the-art zero-shot performance. Achieving SOTA on the vast majority of metrics, CAF-I reaches an average Macro-F1 of 76.31%, a 4.98% absolute improvement over the strongest prior baseline. This success is attained by its effective simulation of human-like multi-perspective analysis, enhancing detection accuracy and interpretability.
UR - https://openreview.net/forum?id=wVXNpw2amK&referrer=%5BAuthor%20Console%5D(%2Fgroup%3Fid%3Dapnns.org%2FICONIP%2F2025%2FConference%2FAuthors%23your-submissions)
M3 - Conference Proceeding
BT - 32nd International Conference on Neural Information Processing (ICONIP)
ER -