LLM-based Few-Shot Early Rumor Detection with Imitation Agent

Fengzhu Zeng; Qian Shao; Ling Cheng; Wei Gao; Shih-Fen Cheng; Jing Ma; Cheng Niu

doi:10.1145/3770854.3780315

LLM-based Few-Shot Early Rumor Detection with Imitation Agent

Fengzhu Zeng
, Qian Shao
, Ling Cheng
, Wei Gao
, Shih-Fen Cheng
, Jing Ma
, Cheng Niu

Department of Computer Science

Research output: Chapter in book/report/conference proceeding › Conference proceeding › peer-review

1 Citation (Scopus)

Abstract

Early Rumor Detection (EARD) aims to identify the earliest point at which a claim can be accurately classified based on a sequence of social media posts. This is especially challenging in data-scarce settings. While Large Language Models (LLMs) perform well in few-shot NLP tasks, they are not well-suited for time-series data and are computationally expensive for both training and inference. In this work, we propose a novel EARD framework that combines an autonomous agent and an LLM-based detection model, where the agent acts as a reliable decision-maker for early time point determination, while the LLM serves as a powerful rumor detector. This approach offers the first solution for few-shot EARD, necessitating only the training of a lightweight agent and allowing the LLM to remain training-free. Extensive experiments on four real-world datasets show our approach boosts performance across LLMs and surpasses existing EARD methods in accuracy and earliness.

Original language	English
Title of host publication	Proceedings of the 32nd ACM SIGKDD Conference on Knowledge Discovery and Data Mining, KDD 2026
Place of Publication	New York
Publisher	Association for Computing Machinery (ACM)
Pages	1856–1867
Number of pages	12
Volume	1
ISBN (Electronic)	9798400722585
DOIs	https://doi.org/10.1145/3770854.3780315
Publication status	Published - Aug 2026
Event	32nd ACM SIGKDD Conference on Knowledge Discovery and Data Mining - International Convention Center Jeju, Jeju Island, Korea, Republic of Duration: 9 Aug 2026 → 13 Aug 2026 https://dl.acm.org/doi/proceedings/10.1145/3770854 (Conference proceeding) https://kdd2026.kdd.org/ (Conference website)

Publication series

Name	Proceedings of the ACM SIGKDD International Conference on Knowledge Discovery and Data Mining
ISSN (Print)	2154-817X

Conference

Conference	32nd ACM SIGKDD Conference on Knowledge Discovery and Data Mining
Abbreviated title	KDD 2026
Country/Territory	Korea, Republic of
City	Jeju Island
Period	9/08/26 → 13/08/26
Internet address	https://dl.acm.org/doi/proceedings/10.1145/3770854 (Conference proceeding) https://kdd2026.kdd.org/ (Conference website)

UN SDGs

This output contributes to the following UN Sustainable Development Goals (SDGs)

SDG 16 Peace, Justice and Strong Institutions

User-Defined Keywords

early rumor detection
few-shot
imitation learning
llms

Access to Document

10.1145/3770854.3780315Licence: CC BY

Cite this

Zeng, F., Shao, Q., Cheng, L., Gao, W., Cheng, S.-F., Ma, J., & Niu, C. (2026). LLM-based Few-Shot Early Rumor Detection with Imitation Agent. In Proceedings of the 32nd ACM SIGKDD Conference on Knowledge Discovery and Data Mining, KDD 2026 (Vol. 1, pp. 1856–1867). (Proceedings of the ACM SIGKDD International Conference on Knowledge Discovery and Data Mining). Association for Computing Machinery (ACM). https://doi.org/10.1145/3770854.3780315

Zeng, Fengzhu ; Shao, Qian ; Cheng, Ling et al. / LLM-based Few-Shot Early Rumor Detection with Imitation Agent. Proceedings of the 32nd ACM SIGKDD Conference on Knowledge Discovery and Data Mining, KDD 2026. Vol. 1 New York : Association for Computing Machinery (ACM), 2026. pp. 1856–1867 (Proceedings of the ACM SIGKDD International Conference on Knowledge Discovery and Data Mining).

@inproceedings{2459d5e26f724a2fa1c5a2665a832012,

title = "LLM-based Few-Shot Early Rumor Detection with Imitation Agent",

abstract = "Early Rumor Detection (EARD) aims to identify the earliest point at which a claim can be accurately classified based on a sequence of social media posts. This is especially challenging in data-scarce settings. While Large Language Models (LLMs) perform well in few-shot NLP tasks, they are not well-suited for time-series data and are computationally expensive for both training and inference. In this work, we propose a novel EARD framework that combines an autonomous agent and an LLM-based detection model, where the agent acts as a reliable decision-maker for early time point determination, while the LLM serves as a powerful rumor detector. This approach offers the first solution for few-shot EARD, necessitating only the training of a lightweight agent and allowing the LLM to remain training-free. Extensive experiments on four real-world datasets show our approach boosts performance across LLMs and surpasses existing EARD methods in accuracy and earliness.",

keywords = "early rumor detection, few-shot, imitation learning, llms",

author = "Fengzhu Zeng and Qian Shao and Ling Cheng and Wei Gao and Shih-Fen Cheng and Jing Ma and Cheng Niu",

note = "This research/project is supported by the National Research Foundation, Singapore under its AI Singapore Programme (AISG Award No: AISG3-RP-2024-035) Publisher Copyright: {\textcopyright} 2026 Copyright held by the owner/author(s); 32nd ACM SIGKDD Conference on Knowledge Discovery and Data Mining, KDD 2026 ; Conference date: 09-08-2026 Through 13-08-2026",

year = "2026",

month = aug,

doi = "10.1145/3770854.3780315",

language = "English",

volume = "1",

series = "Proceedings of the ACM SIGKDD International Conference on Knowledge Discovery and Data Mining",

publisher = "Association for Computing Machinery (ACM)",

pages = "1856–1867",

booktitle = "Proceedings of the 32nd ACM SIGKDD Conference on Knowledge Discovery and Data Mining, KDD 2026",

address = "United States",

url = "https://dl.acm.org/doi/proceedings/10.1145/3770854, https://kdd2026.kdd.org/",

}

Zeng, F, Shao, Q, Cheng, L, Gao, W, Cheng, S-F, Ma, J & Niu, C 2026, LLM-based Few-Shot Early Rumor Detection with Imitation Agent. in Proceedings of the 32nd ACM SIGKDD Conference on Knowledge Discovery and Data Mining, KDD 2026. vol. 1, Proceedings of the ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, Association for Computing Machinery (ACM), New York, pp. 1856–1867, 32nd ACM SIGKDD Conference on Knowledge Discovery and Data Mining, Jeju Island, Korea, Republic of, 9/08/26. https://doi.org/10.1145/3770854.3780315

LLM-based Few-Shot Early Rumor Detection with Imitation Agent. / Zeng, Fengzhu; Shao, Qian; Cheng, Ling et al.
Proceedings of the 32nd ACM SIGKDD Conference on Knowledge Discovery and Data Mining, KDD 2026. Vol. 1 New York: Association for Computing Machinery (ACM), 2026. p. 1856–1867 (Proceedings of the ACM SIGKDD International Conference on Knowledge Discovery and Data Mining).

Research output: Chapter in book/report/conference proceeding › Conference proceeding › peer-review

TY - GEN

T1 - LLM-based Few-Shot Early Rumor Detection with Imitation Agent

AU - Zeng, Fengzhu

AU - Shao, Qian

AU - Cheng, Ling

AU - Gao, Wei

AU - Cheng, Shih-Fen

AU - Ma, Jing

AU - Niu, Cheng

N1 - This research/project is supported by the National Research Foundation, Singapore under its AI Singapore Programme (AISG Award No: AISG3-RP-2024-035) Publisher Copyright: © 2026 Copyright held by the owner/author(s)

PY - 2026/8

Y1 - 2026/8

N2 - Early Rumor Detection (EARD) aims to identify the earliest point at which a claim can be accurately classified based on a sequence of social media posts. This is especially challenging in data-scarce settings. While Large Language Models (LLMs) perform well in few-shot NLP tasks, they are not well-suited for time-series data and are computationally expensive for both training and inference. In this work, we propose a novel EARD framework that combines an autonomous agent and an LLM-based detection model, where the agent acts as a reliable decision-maker for early time point determination, while the LLM serves as a powerful rumor detector. This approach offers the first solution for few-shot EARD, necessitating only the training of a lightweight agent and allowing the LLM to remain training-free. Extensive experiments on four real-world datasets show our approach boosts performance across LLMs and surpasses existing EARD methods in accuracy and earliness.

AB - Early Rumor Detection (EARD) aims to identify the earliest point at which a claim can be accurately classified based on a sequence of social media posts. This is especially challenging in data-scarce settings. While Large Language Models (LLMs) perform well in few-shot NLP tasks, they are not well-suited for time-series data and are computationally expensive for both training and inference. In this work, we propose a novel EARD framework that combines an autonomous agent and an LLM-based detection model, where the agent acts as a reliable decision-maker for early time point determination, while the LLM serves as a powerful rumor detector. This approach offers the first solution for few-shot EARD, necessitating only the training of a lightweight agent and allowing the LLM to remain training-free. Extensive experiments on four real-world datasets show our approach boosts performance across LLMs and surpasses existing EARD methods in accuracy and earliness.

KW - early rumor detection

KW - few-shot

KW - imitation learning

KW - llms

UR - https://dl.acm.org/doi/10.1145/3770854.3780315

UR - https://www.scopus.com/pages/publications/105038079086

U2 - 10.1145/3770854.3780315

DO - 10.1145/3770854.3780315

M3 - Conference proceeding

VL - 1

T3 - Proceedings of the ACM SIGKDD International Conference on Knowledge Discovery and Data Mining

SP - 1856

EP - 1867

BT - Proceedings of the 32nd ACM SIGKDD Conference on Knowledge Discovery and Data Mining, KDD 2026

PB - Association for Computing Machinery (ACM)

CY - New York

T2 - 32nd ACM SIGKDD Conference on Knowledge Discovery and Data Mining

Y2 - 9 August 2026 through 13 August 2026

ER -

Zeng F, Shao Q, Cheng L, Gao W, Cheng SF, Ma J et al. LLM-based Few-Shot Early Rumor Detection with Imitation Agent. In Proceedings of the 32nd ACM SIGKDD Conference on Knowledge Discovery and Data Mining, KDD 2026. Vol. 1. New York: Association for Computing Machinery (ACM). 2026. p. 1856–1867. (Proceedings of the ACM SIGKDD International Conference on Knowledge Discovery and Data Mining). doi: 10.1145/3770854.3780315

LLM-based Few-Shot Early Rumor Detection with Imitation Agent

Abstract

Publication series

Conference

UN SDGs

User-Defined Keywords

Access to Document

Other files and links

Fingerprint

Cite this