license: cc-by-nc-4.0
gated: true
extra_gated_fields:
Full name: text
Affiliation: text
Official email: text
Intended use: text
Country: country
Commercial use:
type: select
options:
- 'No'
- 'Yes'
I agree to the terms: checkbox
| Dataset | Domain | #Tweet | Start time | End time |
|---|---|---|---|---|
| Xinjiang Cotton | Society | 14232 | Mar 15, 2020 | Sep 15, 2020 |
| Moon Landing | Technology | 6721 | Nov 01, 2022 | Nov 01, 2024 |
| Trump-Russia | Politics | 10335 | May 10, 2017 | Nov 10, 2017 |
| Taiwan-Parade | Military | 20746 | Aug 14, 2025 | Sep 08, 2025 |
| Euthanasia | Ethics | 9663 | Jun 23, 2025 | Jul 25, 2025 |
Xinjiang Cotton involves the geopolitical and human rights controversy over forced labor in Xinjiang cotton production chain. The timeline spans from March 15, 2020, to September 15, 2020, to encompass a complete cycle of the controversy. Retrieval tags include
#Xinjiang cotton controversy,#Xinjiang cotton is produced through forced labor, and#Xinjiang labor camps.Moon Landing focuses on the long-standing conspiracy theory claiming that the 1969 Apollo 11 moon landing was staged by NASA. Since it lacks a single acute trigger, we adopt a broader window from November 1, 2022 to November 1, 2024, to observe its persistence and periodic resurgences. Retrieval tags include
#The 1969 moon landing is a HOAX,#NASA faked moon landing, and#the moon landing in 1969 were faked.Trump-Russia centers on the political investigation regarding alleged Russian interference in the 2016 US presidential election. We monitor the period from May 10, 2017, to November 10, 2017, effectively capturing the immediate reaction and subsequent propagation. Retrieval tags include
#Russia intervenes in the US election,#Mueller investigation into the US election, and#Trump-Russia investigation.Taiwan-Parade captures Taiwan's stance on China's 2025 Victory Day military parade. We set the observation window from August 14 to September 8, 2025, to cover the peak discussion period. Retrieval tags include
#Taiwan and China's Victory Day parade,#Taiwan and 93rd Military Parade, and#Taiwanese Officials' Attendance at China's Parade. Notably, as this event occurs after the July 2024 knowledge cutoff of GPT-4o-mini, it is specifically selected to circumvent potential data leakage issues during model evaluation.Euthanasia represents a profound ethical, legal, and medical debate concerning physician-assisted suicide and the right to die with dignity. Recognizing this as a protracted issue characterized by consistent public discourse, we delineated a temporal window spanning June 23 to July 25, 2025. Retrieval tags include
#Euthanasia,#Physician-assisted suicide, and#Death with dignity.
Social media inevitably contains substantial noise, such as duplicated content, advertisements, and posts with weak relevance to the target event. Such noise can severely distort the observed attitude dynamics and undermine data fidelity. To address this, we implement a rigorous filtering pipeline. First, we remove exact duplicates and bot-generated near-duplicates to prevent skew in the attitude distribution. Second, we exclude tweets containing excessively short texts, advertisements, or irrelevant external links to preserve substantive semantic content. Third, we exclude accounts exhibiting abnormal bot-like behaviors, such as posting more than 100 tweets per day, thereby better approximating realistic human social interactions. In addition, the collected data are restricted to English-language tweets to ensure linguistic consistency. Finally, we employ an LLM-based semantic filter to remove tweets with low semantic relevance to the target event.
Guided by the principle of open-source collaboration, we promote this dataset as an out-of-the-box benchmark for evaluating large-scale social networks simulators. Accordingly, \textit{DynamiX} have established a systematic evaluation pipeline spanning several core dimensions, including macro alignment evaluation, large-scale collective behavior analysis, stability analysis and intervention strategies. We anticipate that this datasets will serve as a valuable resource for the community and facilitate further advances in both multi-agent systems and social network simulation research.
@misc{sun2025dynamixlargescaledynamicsocial, title={DynamiX: Large-Scale Dynamic Social Network Simulator}, author={Yanhui Sun and Wu Liu and Wentao Wang and Hantao Yao and Jiebo Luo and Yongdong Zhang}, year={2025} }