Zhiyu Zoey Chen
@ZhiyuChen4
Followers
2K
Following
841
Media
14
Statuses
175
NLP researcher. Assistant Professor @UT_Dallas. Postdoc @CarnegieMellon. Ph.D. @UCSBCS. #NLProc.
Dallas, TX
Joined May 2018
I'm shocked to see racism happening in academia again, at the best AI conference @NeurIPSConf. Targeting specific ethnic groups to describe misconduct is inappropriate and unacceptable. @NeurIPSConf must take a stand. We call on Rosalind Picard @MIT @medialab to retract and
142
296
2K
I will be joining @UT_Dallas as an Assistant Professor in @UTDJonsson in Fall 2024. Deep thanks to my advisors @WilliamWangNLP, Xifeng Yan, collaborators, and friends for supporting me as always. Currently, I'm working as a postdoc in @S3DatCMU. Look forward to the new journey!.
29
8
251
🧠🤖 Can large language models provide assistance in psychotherapy?. Our #EMNLP2023 findings paper showcases how #LLM can effectively help diagnose distorted thoughts for cognitive behavior therapy (CBT). Excited for the future of #AI and #psychotherapy in
4
40
191
Our group at Meta Reality Labs is hiring Ph.D. Research Interns to work on various Multimodal & NLP related projects. We aim at publishing in top-tier conferences such as CVPR, ACL, EMNLP, ICLR, etc. Feel free to reach out (DM/email) if interested.
We are hiring PhD Research Interns to work on various Multimodal & NLP related projects (Reality Labs) for 2023. See JDs here -- apply directly or reach out to me directly via email! .- -
2
12
97
💡Introducing Patient-Ψ: 🧠🤖 We integrate cognitive modeling with Large Language models to simulate patients with mental health disorders, creating a new framework for training professionals. 📄 Paper: Based on a user study of 33 mental health experts
3
13
87
Official Dr. Chen now. Thanks to all who ever offered me help. Hope I can also change things, and help others in the future.
6
1
85
📈🏥⚖️How to build LLM applications in vital sectors such as finance, healthcare, and law? . 📜Arxiv: . Sharing our new survey paper: A Survey on Large Language Models for Critical Societal Domains: Finance, Healthcare, and Law. We explore the LLM research
5
15
71
Glad to share our ConvFinQA paper accepted to @emnlp2022 #EMNLP2022. We create a new dataset exploring the chains of numerical reasoning in conversational finance questions answering. #NLProc.Paper: Data & code: (1/3)
2
8
61
2022.6 PhD in Computer Science! Honored to be hooded by Prof. William Wang @WilliamWangNLP and Prof. Xifeng Yan
2
0
61
🚀🚀 Excited to share that our paper "A Survey on Large Language Models for Critical Societal Domains: Finance, Healthcare, and Law" has been accepted by @TmlrOrg #TMLR with a Survey Certification! . 🔗Paper: Congrats to the great team: @JingMa77838617,.
📈🏥⚖️How to build LLM applications in vital sectors such as finance, healthcare, and law? . 📜Arxiv: . Sharing our new survey paper: A Survey on Large Language Models for Critical Societal Domains: Finance, Healthcare, and Law. We explore the LLM research
4
14
58
My amazing coauthor @steph_milani will present our work at @NeurIPSConf!! Stop by to check how to build LLM applications for mental health training.
🇨🇦 Hi! I’m attending my last @NeurIPSConf as a PhD student, presenting Patient-Ψ at a few workshops. I'm on the job market, looking for TT faculty roles & post-docs. DM if you'd like to chat (or invite me to a party 🥳)!
0
6
48
In our #EMNLP2023 paper, we demonstrate LLMs can effectively analyze maladaptive thought patterns from mental health issues, as to assist cognitive behavior therapy. Curious to find out to what extent language can represent human psychological constructs and cognition functions.
Using large language models in psychology: . 💡 LLMs have the potential to advance psychological measurement, experimentation and practice. 💡 LLM generated on-topic, grammatically correct useless information, but not based on research and psychology construct. 💡A critical
0
10
40
Our work Patient-Ψ has been accepted to #EMNLP2024 Main! 🎉 .- We use LLMs to simulate mental health patients, and our user study with 33 professionals shows great results. - Huge kudos to @RuiyiWang153 and @steph_milani for leading this effort! See you in Miami 🏖️.
(1/n) Can large language models simulate patients with mental health conditions?. We introduce Patient-Ψ🤖, where we integrate cognitive modeling with LLMs to simulate patients for training mental health professionals. 📎Paper link:
3
5
39
🤖 Can LLMs effectively assist in cognitive behavior therapy (CBT)?. 🔗New paper: We present the first systematic benchmark to evaluate LLMs' efficacy for CBT. We include three levels of tasks: basic CBT knowledge acquisition, cognitive model.
📢 Excited to introduce CBT-Bench (, a new benchmark that systematically evaluates LLMs’ capabilities in Cognitive Behavioral Therapy (CBT) across three Levels:.
0
6
37
Happy to share that our paper "KETOD: Knowledge-Enriched Task-Oriented Dialogue" was accepted to #NAACL2022 Findings! w/ Bing Liu, @shane_moon, @Chinnadhurai @pcrook_academic @WilliamWangNLP .Code and data will be out soon. #NLProc
0
6
35
Excited to share that our #EMNLP2024 paper on patient simulation received the Best Paper Award at the GenAI4Health workshop @NeurIPSConf! 🎉 Join @steph_milani's talk and explore Patient-Ψ!. 📍 12/14 | East Meeting Room 16.🕙 10:40am - Oral Presentation.🕐 1:00pm - Poster Session.
🇨🇦 Hi! I’m attending my last @NeurIPSConf as a PhD student, presenting Patient-Ψ at a few workshops. I'm on the job market, looking for TT faculty roles & post-docs. DM if you'd like to chat (or invite me to a party 🥳)!
0
2
34
🚀Our work Patient-Ψ has been featured in CMU SCS news! Check out how we use LLM to simulate patients for mental health training:
(1/n) Can large language models simulate patients with mental health conditions?. We introduce Patient-Ψ🤖, where we integrate cognitive modeling with LLMs to simulate patients for training mental health professionals. 📎Paper link:
1
3
32
The 1st Workshop on Robust NLP for Finance (RobustFin) at #KDD2023 is welcoming submissions and shared task participations! Papers are due by May 23rd. Shared task due by June 18 with prizes. Details are available at:
3
11
30
Check out our new work on rule-learning of LLM agents🤖 in interactive environments!. We propose IDEA, a holistic LLM agent framework integrating Induction, DEduction, and Abduction, to generate hypotheses, devise plans, and revise hypotheses iteratively to learn new rules in.
(1/n) 🧠💡Humans learn new rules through iterative reasoning: abduction, induction, and deduction. Can LLM agents do the same?.Introducing RULEARN—a benchmark to assess LLMs' rule-learning abilities in interactive environments. We propose IDEA, a holistic framework integrating
1
6
30
We just released our KETOD dataset. Check it out here at
Happy to share that our paper "KETOD: Knowledge-Enriched Task-Oriented Dialogue" was accepted to #NAACL2022 Findings! w/ Bing Liu, @shane_moon, @Chinnadhurai @pcrook_academic @WilliamWangNLP .Code and data will be out soon. #NLProc
0
5
28
We have reached the end of the FinQA challenge of the #sukiworkshop at #naacl2022. Now we announce the champion from the Ant Group, the runner-up from Nanjing University, and two honorable mentions from Google Research and LYZD-FinTech. Congratulations to all the winners!.
1
2
21
Check out our new #EMNLP2024 paper on multimodal procedural planning!.
Excited to share that our paper is in EMNLP 2024 Findings! We received positive reviews (4/3.5/3.5) and a tough critique from AC. Thanks to the strong rebuttal, we made it through. Looking forward to meeting old friends and new ones in Miami!. 📜paper:
2
3
20
Large language models can simulate communicative behaviors based on cognitive models. Check out our new work to see how it mimics the way patients with mental health disorders speak.
(1/n) Can large language models simulate patients with mental health conditions?. We introduce Patient-Ψ🤖, where we integrate cognitive modeling with LLMs to simulate patients for training mental health professionals. 📎Paper link:
0
1
16
Check out our new #ACL2024 paper on zero-shot dialogue state tracking using function calling!.
Thrilled to announce our work is accepted to #ACL2024 Main! We’re the first to solve zero-shot DST with Function Call/Tool Use, bridging the gap and achieving remarkable performance with both 7b/13b OSS models and GPT-3.5/4. More results coming soon. Code:
1
1
15
Stephanie is one of the top candidates you don’t want to miss! She has the rare combination of exceptional technical talent and the creativity, communication skills, and empathy essential for human-centered research. Working with her has been inspiring—I’ve learned so much.
🎇 I’m on the academic job market! I’m a PhD candidate at @mldcmu. My research tackles challenges that arise from the sequential nature of human-AI interaction. Toward this goal, my work involves:.🤖 reinforcement learning, .🧠 foundation models, and .👩💻 human-centered AI.
1
0
14
How instruction fine-tuning with code can boost LLMs’ reasoning abilities? Check out our recent work ⬇️.
🚀 Excited to share our latest research on investigating the effect of coding data on LLMs' reasoning abilities! 💻🔍 Discover how Instruction Fine-Tuning with code can boost zero-shot performance across various tasks and domains. 📊 🔗
1
2
13
📣 Stop by Nov 12 at 11:00 (Poster Session A) to play with Patient-Ψ with my amazing coauthor @RuiyiWang153! #EMNLP2024.
Heading to Miami 🏝️ Looking forward to meeting the amazing researchers at #EMNLP2024!. If you’re interested in LLM agents, social intelligence, mental health, …, plz message me on Whova app!. I’ll present Patient-Ψ on Nov 12 at 11:00 (Poster Session A) — hope to see you there!
0
1
13
📣We've released the code and data for our IDEA project! Dive in and explore here: 🔗 Happy playing! 🚀.
Check out our new work on rule-learning of LLM agents🤖 in interactive environments!. We propose IDEA, a holistic LLM agent framework integrating Induction, DEduction, and Abduction, to generate hypotheses, devise plans, and revise hypotheses iteratively to learn new rules in.
0
1
11
Check out our new work using function calling for dialogue state tracking. We set up new SOTA and train Llama 13B chat model comparable to ChatGPT.
Thanks for sharing our work! . 🚀 FnCTOD: a method that empowers chat-based LLMs with function-calling abilities, enabling them to handle complex, task-oriented conversations through the appropriate use of tools and API calls (DST, a type of database search API). 🧵 [1/n].
0
0
10
We would like to thank all the teams participating in our challenge. Looking forward to seeing you at #naacl2022. @WilliamWangNLP @ruizhang_nlp @WenhuChen @sameenashah_AI @windx0303.
0
0
9
The Valdez glacier lake. Best tour ever. Want to stay in this ice cave, read, and meditate.
0
0
9
Our SUKI workshop will be held at #NAACL2022, submission is welcome: Please consider participating in our FinQA shared task: The top winners will be awarded cash prizes sponsored by J.P. Morgan!.
Hello World! Structured and Unstructured Knowledge Integration (SUKI) workshop at #NAACL2022 is welcoming submissions and shared task participations🙌! Papers due by April 8. Two shared tasks due by June 8 with cash awards🥰. Details are available 👉
0
0
9
Well I study NLP, I do programming, and my hobbies are mostly art things 🙃
0
0
8
Michael has done incredible works, he has excellent mentoring skills, and he’s a super fun person to work with. You should definitely hire him!.
🚨😱Obligatory job market announcement post‼️🤯. I'm searching for faculty positions/postdocs in multimodal/multilingual NLP and generative AI!. I'll be at #NeurIPS2024 presenting our work on meta-evaluation for text-to-image faithfulness! Let's chat!. Website in bio, papers in🧵
0
1
8
Love this movie version of Rick & Morty.
Michelle Yeoh won the best actress for her role in “Everything, Everywhere All at Once.” The victory makes her the first Asian star to win best actress in the 95-year history of the Academy Awards.
0
0
8
Our CBT-Bench paper has been accepted to #NAACL2025 Main! Congrats to the lead @_Guuuuuuuu_ and @Qnolan4 . See you in Albuquerque!.
🤖 Can LLMs effectively assist in cognitive behavior therapy (CBT)?. 🔗New paper: We present the first systematic benchmark to evaluate LLMs' efficacy for CBT. We include three levels of tasks: basic CBT knowledge acquisition, cognitive model.
0
2
8
I’ll tell myself “let’s think step by step” every time I confront something hard.
Large Language Models are Zero-Shot Reasoners. Simply adding “Let’s think step by step” before each answer increases the accuracy on MultiArith from 17.7% to 78.7% and GSM8K from 10.4% to 40.7% with GPT-3.
0
0
6
Thanks to the great team @RuiyiWang153, @steph_milani, @imjamiechiu, Jiayin Zhi, Shaun Eack, Travis Labrum, Samuel Murphy, @viscidula, Kate Hardy, @hongshenus, @fangf07.
0
0
6
We find that LLMs can effectively analyze people’s thoughts by distinguishing subjective/objectives, elaborating supporting/contradicting evidence, and summarizing underlying thought schemas. Through these stages, it can predict cognitive distortions based on patients’ speech.
1
0
5
As we navigate the potential of LLM in psychotherapy, we should always take caution for such development. Our foremost priority should remain on building systems that are safe and responsible. The partnership between professionals and AI systems is a promising direction for the.
1
0
5
@emnlp2022 Current models like GPT3 fail drastically on our dataset, posing great challenges in modeling long-range, complex numerical reasoning paths. How to incorporate large language models into such real-world application domains should be one important next research focus. (2/3).
1
0
5
Mental illness remains one of the most critical public health issues of our time, due to the severe scarcity and accessibility limit of professionals. In the era of Large Language Models, we believe it is the right time to develop AI assistance for computational psychotherapy.
1
0
5
A big shoutout to our great team: @JingMa77838617, @XZ1023_, Nan Hao, @AnYan_ai, @arminehnouri, @Qnolan4, Julian McAuley, Linda Petzold, @WilliamWangNLP.
0
0
4
Happy to discuss our work on combining task oriented dialogue and knowledge-enriched chitchat. Joint work w/ Bing Liu, @shane_moon @Chinnadhurai @pacrook @WilliamWangNLP #NLProc.
Stop by 10:45-12:15 Wed morning's #NAACL2022 poster session at Regency A & B to learn about @ZhiyuChen4's Meta AI work on KETOD: Knowledge-Enriched Task-Oriented Dialogue
0
0
4
Check out our new work on multimodal procedural planning! We propose a novel elegant framework to bridge text and visual information, mutually enhancing each other. #NLProc.
🚀Thrilled to release #TIP (Dual Text-Image Prompting), a #DALLE2 #StableDiffusion-2 enhanced #LLM that can generate coherent and authentic multimodal procedural plans toward a high-level goal. 🧵8. 📜paper: 🔗data & code:
0
0
4
Thanks for sharing our work! #LLM will significantly benefit psychotherapy and computational psychiatry. We expect to build powerful LLM agent to assist therapists, to cope with the global mental health crisis and the shortage of professionals.
Diagnosis of Thought (DoT) Prompting is a 3-stage framework that provides step-by-step guidance for an AI system to detect cognitive distortions from a patient's speech. Stage 1 - Subjectivity Assessment: Separate objective facts from subjective thoughts and opinions. Stage 2
0
2
3
@emnlp2022 Thanks to the collaborators @ShiyangLi6, Charese Smiley, Zhiqiang Ma, @sameenashah_AI @WilliamWangNLP (3/3).
0
0
3
@WenhuChen NLP + psychology. I’ve studied both lol. For example, to analyze the mental disorders, trauma, cause of depression, mental process, etc etc, based on people’s speech. Tho most therapists can not even do this really well.
0
0
3
Big congrats to @WilliamWangNLP on launching the world's first AI agent for chip design and verification! Excited to see how it can revolutionize this area.
🚀Introducing ChipAgents: the World's First AI Agent for Chip Design and Verification. Get ready to supercharge your workflow and accelerate your time-to-market! 💻⚡
0
1
3
@WenhuChen Check the black mirror s2 e1 Be right back and s3 e4 San Junipero. You can even think make it multimodal 😂.
1
0
1
We will have the ConvFinQA dataset for our shared task. It’s one of the few conversational QA benchmarks that LLM still struggles with; Domain specific LLMs like BloombergGPT cannot even work very well #NLProc.
0
0
1
0
0
0