Yong Cao Profile
Yong Cao

@YongCaoPlus

Followers
83
Following
112
Media
4
Statuses
31

Postdoc at the University of Tübingen, supervised by @andreasgeiger0 | Research: Scientific Document Analysis, Cultural AI, & LLM Agents

Tübingen, Germany
Joined October 2022
Don't wanna be here? Send us removal request.
@YongCaoPlus
Yong Cao
2 months
Excited to join the IT:U summer school this year! Looking forward to learning, sharing, and connecting with amazing researchers! 😀.
@yufanghou
Yufang Hou
2 months
‼️Call for Participation: IT:U Summer School 2025 on Natural Language Processing‼️. We are excited to announce that the IT:U Summer School 2025, focused on NLP, will take place on July 25–26 in Linz, Austria, just ahead of #ACL2025. 🔗 Register here:
Tweet media one
0
0
2
@YongCaoPlus
Yong Cao
2 months
RT @KatrinRenz: 📣 Excited to share our #CVPR2025 Spotlight paper and my internship project @wayve_ai: SimLingo. A Vision-Language-Action (….
0
13
0
@YongCaoPlus
Yong Cao
3 months
RT @LauraCabelloPi: 📣 Excited about our Cross Cultural Considerations in NLP workshop! @c3_nlp @ @naaclmeeting coming up in no time—just tw….
c3nlp.github.io
0
5
0
@YongCaoPlus
Yong Cao
5 months
For more analysis and details, check out our paper: .Our code is available here:
Tweet media one
github.com
Code for Simulating LLMs' cultural responses distribution. - yongcaoplus/SimLLMCultureDist
0
0
1
@YongCaoPlus
Yong Cao
5 months
🔍 Findings: . Testing on three model families & seven models, our method consistently outperforms other approaches —even on unseen questions, countries, and an entirely new survey. While challenges still remain, we demonstrate the benefits of specialization for simulation.🚀
Tweet media one
0
0
1
@YongCaoPlus
Yong Cao
5 months
📂 Dataset: . We construct the simulation dataset from the World Values Survey (WVS) in both English & Chinese. To test generalization, we use a subset of GlobalOpinionQA, specifically the Pew Global Attitudes Survey.
Tweet media one
0
0
1
@YongCaoPlus
Yong Cao
5 months
🛠 Method: . We introduce a fine-tuning approach that uses first-token probabilities to align predicted and actual survey response distributions. By optimizing with KL divergence loss, our method effectively minimizes divergence from real-world survey data. 📊.
0
0
1
@YongCaoPlus
Yong Cao
5 months
💡Motivation:. Large-scale surveys are crucial for social science, but they’re expensive and time-consuming. What if LLMs could simulate group-level survey results? Thus, we try to specialize LLMs for simulating this task, using country-level data from cultural surveys. 🌍.
0
0
1
@YongCaoPlus
Yong Cao
5 months
🎉 Excited to share our NAACL main paper on simulating survey response distributions via first-token alignment fine-tuning. Many thanks to our collaborators: Haijiang Liu, @rnav_arora @IAugenstein @paul_rottger and @daniel_hers. More details, paper link and code, see below👇
Tweet media one
5
3
16
@YongCaoPlus
Yong Cao
5 months
RT @omarsar0: Transforming Science with LLMs. A comprehensive survey on AI-assisted scientific discovery, experimentation, content generati….
0
84
0
@YongCaoPlus
Yong Cao
5 months
Very interesting work that comprehensively investigates cultural knowledge in LLMs in the domain of food! Looking forward to more discussion at NAACL.
@LiZhou21
Li Zhou
5 months
📢Happy to share our paper "Does Mapo Tofu Contain Coffee? Probing LLMs for Food-related Cultural Knowledge" has been accepted to #NAACL2025 main!.Work with Taelin Karidi, Wanlong Liu, @nGarneau @YongCaoPlus , Wenyu Chen, Haizhou Li, @daniel_hers.
Tweet media one
0
0
3
@YongCaoPlus
Yong Cao
6 months
RT @c3_nlp: 📣 Second Call for Papers! 🎉. We’re extending our CFP to February 6, 2025 ⏳—more time to submit your work on culture, society &….
c3nlp.github.io
0
4
0
@YongCaoPlus
Yong Cao
7 months
It's hugely disappointing to hear such arguments from a top AI conference. 😡More efforts are needed to avoid LLMs aligning with these offensive points. (Sometimes LLMs behave more polite and respectful so far. ).
@drjingjing2026
Jing-Jing Li @ICML
7 months
1/3 Today, an anecdote shared by an invited speaker at #NeurIPS2024 left many Chinese scholars, myself included, feeling uncomfortable. As a community, I believe we should take a moment to reflect on why such remarks in public discourse can be offensive and harmful.
Tweet media one
0
1
2
@YongCaoPlus
Yong Cao
7 months
Looking forward to discussing fascinating cultural research and meeting again next May! 😊. Please check for our CFP details if you are interested!🗓️.
@c3_nlp
C3NLP
7 months
📣 C3NLP 2025 CFP now out 🎉. We’re thrilled to announce C3NLP will be at NAACL 2025 in Albuquerque, New Mexico! 🌎. 📅 Workshop Details: Co-organized w/ @vinodkpg @sunipa17 @LucianaBenotti @daniel_hers @YongCaoPlus @LauraCabelloPi @LiZhou21 @IfeAdebara.
0
0
2
@YongCaoPlus
Yong Cao
8 months
Thrilled to have participated in such an insightful workshop! A fantastic opportunity to meet, brainstorm, and engage in discussions with brilliant minds.
@PontiEdoardo
Edoardo Ponti
8 months
We had a blast at this year's @ELLISforEurope @dagstuhl seminar on "Modular and Agentive LLMs". Thanks everyone for participating!
Tweet media one
Tweet media two
Tweet media three
0
0
6
@YongCaoPlus
Yong Cao
9 months
RT @delliott: Recent work led by @constanzafierro and @negarforoutan on understanding the mechanisms behind multilingual factual recall.
0
3
0
@YongCaoPlus
Yong Cao
11 months
RT @LauraCabelloPi: Amazing brochure to #ACL2024 🇹🇭 with an amazing 2nd edition of the workshop @c3_nlp 🗺️ with amazing co-organizers ♥️ ht….
0
4
0
@YongCaoPlus
Yong Cao
1 year
RT @vinodkpg: Siva is ready for the #acl2024 conference, Aug 11-16! Are you? . He's especially looking forward to the exciting talks & pane….
0
4
0
@YongCaoPlus
Yong Cao
1 year
RT @c3_nlp: Excited about our Cross Cultural Considerations in #NLProc workshop @c3_nlp @ #ACL2024, in just a month’s time!. We'll have 25+….
0
9
0
@YongCaoPlus
Yong Cao
1 year
RT @vinodkpg: Signal boosting this financial support for registration for the @c3_nlp workshop at #acl2024! Hope to see you in Bangkok, in….
0
3
0