Yuxuan Li
@yuxuanli1225.bsky.social
Second-year PhD student @CMU HCII.
Previously undergrad @Tsinghua CS and research intern @Berkeley AI Research @Berkeley InfoSchool.
https://yuxuanli.com/
Previously undergrad @Tsinghua CS and research intern @Berkeley AI Research @Berkeley InfoSchool.
https://yuxuanli.com/
Excited to share that our paper has been accepted to 𝗜𝗖𝗠𝗟 𝟮𝟬𝟮𝟲! 🎉
Multi-agent is everywhere today. But put frontier LLMs in a room where each holds a different piece of the puzzle, and they fail 70% of the time.
Here's why:
📄 Paper: arxiv.org/abs/2505.11556
Multi-agent is everywhere today. But put frontier LLMs in a room where each holds a different piece of the puzzle, and they fail 70% of the time.
Here's why:
📄 Paper: arxiv.org/abs/2505.11556
Systematic Failures in Collective Reasoning under Distributed Information in Multi-Agent LLMs
Multi-agent systems built on large language models (LLMs) are expected to enhance decision-making by pooling distributed information, yet systematically evaluating this capability has remained challen...
arxiv.org
April 30, 2026 at 6:55 PM
Excited to share that our paper has been accepted to 𝗜𝗖𝗠𝗟 𝟮𝟬𝟮𝟲! 🎉
Multi-agent is everywhere today. But put frontier LLMs in a room where each holds a different piece of the puzzle, and they fail 70% of the time.
Here's why:
📄 Paper: arxiv.org/abs/2505.11556
Multi-agent is everywhere today. But put frontier LLMs in a room where each holds a different piece of the puzzle, and they fail 70% of the time.
Here's why:
📄 Paper: arxiv.org/abs/2505.11556
Excited to share that our paper has been accepted to ICML 2026! 🎉
📄 Paper: arxiv.org/abs/2505.11556
📊 Benchmark: huggingface.co/datasets/Yux...
#ICML
📄 Paper: arxiv.org/abs/2505.11556
📊 Benchmark: huggingface.co/datasets/Yux...
#ICML
April 30, 2026 at 6:42 PM
Excited to share that our paper has been accepted to ICML 2026! 🎉
📄 Paper: arxiv.org/abs/2505.11556
📊 Benchmark: huggingface.co/datasets/Yux...
#ICML
📄 Paper: arxiv.org/abs/2505.11556
📊 Benchmark: huggingface.co/datasets/Yux...
#ICML
The first PoliSim workshop at #CHI2026 was a huge success! We saw incredible interest from researchers across diverse domains, with attendance nearly filling one of the largest rooms at the venue.
1/n
1/n
April 19, 2026 at 7:06 PM
The first PoliSim workshop at #CHI2026 was a huge success! We saw incredible interest from researchers across diverse domains, with attendance nearly filling one of the largest rooms at the venue.
1/n
1/n
Excited to be heading to Barcelona for #CHI2026 to host our workshop PoliSim: LLM Agent Simulation for Policy!
This year, we’ve seen incredible interest from researchers across HCI, NLP, CSS, and Policy. We accepted 25 outstanding papers, with 5 selected as Best Paper nominees.
This year, we’ve seen incredible interest from researchers across HCI, NLP, CSS, and Policy. We accepted 25 outstanding papers, with 5 selected as Best Paper nominees.
April 11, 2026 at 5:38 PM
Excited to be heading to Barcelona for #CHI2026 to host our workshop PoliSim: LLM Agent Simulation for Policy!
This year, we’ve seen incredible interest from researchers across HCI, NLP, CSS, and Policy. We accepted 25 outstanding papers, with 5 selected as Best Paper nominees.
This year, we’ve seen incredible interest from researchers across HCI, NLP, CSS, and Policy. We accepted 25 outstanding papers, with 5 selected as Best Paper nominees.
Can LLMs really serve as "crash dummies" for security & privacy testing? We put this assumption to the test.
🚨New preprint 🚨: "How Well Can LLM Agents Simulate End-User Security and Privacy Attitudes and Behaviors?"
👇 THREAD 👇
[Link to paper: arxiv.org/abs/2602.184...
[1/n]
🚨New preprint 🚨: "How Well Can LLM Agents Simulate End-User Security and Privacy Attitudes and Behaviors?"
👇 THREAD 👇
[Link to paper: arxiv.org/abs/2602.184...
[1/n]
https://arxiv.org/abs/2602.18464]
February 25, 2026 at 5:47 PM
Can LLMs really serve as "crash dummies" for security & privacy testing? We put this assumption to the test.
🚨New preprint 🚨: "How Well Can LLM Agents Simulate End-User Security and Privacy Attitudes and Behaviors?"
👇 THREAD 👇
[Link to paper: arxiv.org/abs/2602.184...
[1/n]
🚨New preprint 🚨: "How Well Can LLM Agents Simulate End-User Security and Privacy Attitudes and Behaviors?"
👇 THREAD 👇
[Link to paper: arxiv.org/abs/2602.184...
[1/n]
Excited to share that I’m joining @msftresearch.bsky.social AI Frontiers this summer as a Research Intern, working on LLM Agents, and reporting to @zacharyhuang.bsky.social. I’ll be based in Redmond — would love to connect with folks around Seattle!
February 18, 2026 at 5:32 PM
Excited to share that I’m joining @msftresearch.bsky.social AI Frontiers this summer as a Research Intern, working on LLM Agents, and reporting to @zacharyhuang.bsky.social. I’ll be based in Redmond — would love to connect with folks around Seattle!
Reposted by Yuxuan Li
🚨New paper🚨
We spent a year working with emergency preparedness policymakers to answer a simple question: can LLM agent simulations actually help real institutions make better decisions?
The answer is yes—but perhaps not how you'd expect.
👇 THREAD 👇
[Link to paper: arxiv.org/abs/2509.218...
[1/n]
We spent a year working with emergency preparedness policymakers to answer a simple question: can LLM agent simulations actually help real institutions make better decisions?
The answer is yes—but perhaps not how you'd expect.
👇 THREAD 👇
[Link to paper: arxiv.org/abs/2509.218...
[1/n]
February 13, 2026 at 6:42 PM
🚨New paper🚨
We spent a year working with emergency preparedness policymakers to answer a simple question: can LLM agent simulations actually help real institutions make better decisions?
The answer is yes—but perhaps not how you'd expect.
👇 THREAD 👇
[Link to paper: arxiv.org/abs/2509.218...
[1/n]
We spent a year working with emergency preparedness policymakers to answer a simple question: can LLM agent simulations actually help real institutions make better decisions?
The answer is yes—but perhaps not how you'd expect.
👇 THREAD 👇
[Link to paper: arxiv.org/abs/2509.218...
[1/n]
🚨New paper🚨
We spent a year working with emergency preparedness policymakers to answer a simple question: can LLM agent simulations actually help real institutions make better decisions?
The answer is yes—but perhaps not how you'd expect.
👇 THREAD 👇
[Link to paper: arxiv.org/abs/2509.218...
[1/n]
We spent a year working with emergency preparedness policymakers to answer a simple question: can LLM agent simulations actually help real institutions make better decisions?
The answer is yes—but perhaps not how you'd expect.
👇 THREAD 👇
[Link to paper: arxiv.org/abs/2509.218...
[1/n]
February 13, 2026 at 6:42 PM
🚨New paper🚨
We spent a year working with emergency preparedness policymakers to answer a simple question: can LLM agent simulations actually help real institutions make better decisions?
The answer is yes—but perhaps not how you'd expect.
👇 THREAD 👇
[Link to paper: arxiv.org/abs/2509.218...
[1/n]
We spent a year working with emergency preparedness policymakers to answer a simple question: can LLM agent simulations actually help real institutions make better decisions?
The answer is yes—but perhaps not how you'd expect.
👇 THREAD 👇
[Link to paper: arxiv.org/abs/2509.218...
[1/n]
🚨 Deadline Extended!
Due to popular demand, the submission deadline for the PoliSim workshop at CHI’26 has been extended to February 20th! 🎉
We look forward to your submissions!
👉 polisim.net
Due to popular demand, the submission deadline for the PoliSim workshop at CHI’26 has been extended to February 20th! 🎉
We look forward to your submissions!
👉 polisim.net
February 11, 2026 at 10:22 PM
🚨 Deadline Extended!
Due to popular demand, the submission deadline for the PoliSim workshop at CHI’26 has been extended to February 20th! 🎉
We look forward to your submissions!
👉 polisim.net
Due to popular demand, the submission deadline for the PoliSim workshop at CHI’26 has been extended to February 20th! 🎉
We look forward to your submissions!
👉 polisim.net
Reposted by Yuxuan Li
📣 New at #CHI2026
Developing a new AI product? How would you figure out what are the privacy risks?
Privy help non-privacy expert practitioners create high quality privacy impact assessments for early-stage AI products.
Led by @hankhplee.bsky.social
Paper: www.sauvik.me/papers/69/s...
Developing a new AI product? How would you figure out what are the privacy risks?
Privy help non-privacy expert practitioners create high quality privacy impact assessments for early-stage AI products.
Led by @hankhplee.bsky.social
Paper: www.sauvik.me/papers/69/s...
February 9, 2026 at 7:12 PM
📣 New at #CHI2026
Developing a new AI product? How would you figure out what are the privacy risks?
Privy help non-privacy expert practitioners create high quality privacy impact assessments for early-stage AI products.
Led by @hankhplee.bsky.social
Paper: www.sauvik.me/papers/69/s...
Developing a new AI product? How would you figure out what are the privacy risks?
Privy help non-privacy expert practitioners create high quality privacy impact assessments for early-stage AI products.
Led by @hankhplee.bsky.social
Paper: www.sauvik.me/papers/69/s...
Don’t forget! The submission deadline for our CHI 2026 workshop "PoliSim: LLM Agent Simulation for Policy" is only one week away (Feb. 13). We’d love to see your work!
🚨 New CHI 2026 Workshop 🚨
PoliSim@CHI 2026: LLM Agent Simulation for Policy
PoliSim@CHI 2026: LLM Agent Simulation for Policy
February 5, 2026 at 6:41 PM
Don’t forget! The submission deadline for our CHI 2026 workshop "PoliSim: LLM Agent Simulation for Policy" is only one week away (Feb. 13). We’d love to see your work!
🚨 New CHI 2026 Workshop 🚨
PoliSim@CHI 2026: LLM Agent Simulation for Policy
PoliSim@CHI 2026: LLM Agent Simulation for Policy
December 18, 2025 at 4:26 PM
🚨 New CHI 2026 Workshop 🚨
PoliSim@CHI 2026: LLM Agent Simulation for Policy
PoliSim@CHI 2026: LLM Agent Simulation for Policy
Reposted by Yuxuan Li
🚀💫 I’m on the job market for academic (tenure-track) and industry research positions!
👋I am a Postdoc Fellow at @hcii.cmu.edu working at the intersection of human-AI interaction, cognitive science, responsible AI, design, and social computing. I earned my PhD from @gtresearch.bsky.social in 2024.
👋I am a Postdoc Fellow at @hcii.cmu.edu working at the intersection of human-AI interaction, cognitive science, responsible AI, design, and social computing. I earned my PhD from @gtresearch.bsky.social in 2024.
November 11, 2025 at 11:13 PM
🚀💫 I’m on the job market for academic (tenure-track) and industry research positions!
👋I am a Postdoc Fellow at @hcii.cmu.edu working at the intersection of human-AI interaction, cognitive science, responsible AI, design, and social computing. I earned my PhD from @gtresearch.bsky.social in 2024.
👋I am a Postdoc Fellow at @hcii.cmu.edu working at the intersection of human-AI interaction, cognitive science, responsible AI, design, and social computing. I earned my PhD from @gtresearch.bsky.social in 2024.
Happy to share that our paper has been selected for an ORAL presentation at #EMNLP2025 main conference!
I’ll be presenting in person in Suzhou. See you there!
I’ll be presenting in person in Suzhou. See you there!
🚨 New Preprint! 🚨
Smarter LLMs are more selfish.
We show reasoning-enhanced models significantly prefer greed over cooperation. The more LLMs reason, the worse they cooperate.
👇 THREAD 👇
[Link to paper: arxiv.org/abs/2502.177...
[1/n]
Smarter LLMs are more selfish.
We show reasoning-enhanced models significantly prefer greed over cooperation. The more LLMs reason, the worse they cooperate.
👇 THREAD 👇
[Link to paper: arxiv.org/abs/2502.177...
[1/n]
September 15, 2025 at 2:44 PM
Happy to share that our paper has been selected for an ORAL presentation at #EMNLP2025 main conference!
I’ll be presenting in person in Suzhou. See you there!
I’ll be presenting in person in Suzhou. See you there!
Happy to share that our paper has been accepted to #EMNLP2025 main conference!
🚨 New Preprint! 🚨
Smarter LLMs are more selfish.
We show reasoning-enhanced models significantly prefer greed over cooperation. The more LLMs reason, the worse they cooperate.
👇 THREAD 👇
[Link to paper: arxiv.org/abs/2502.177...
[1/n]
Smarter LLMs are more selfish.
We show reasoning-enhanced models significantly prefer greed over cooperation. The more LLMs reason, the worse they cooperate.
👇 THREAD 👇
[Link to paper: arxiv.org/abs/2502.177...
[1/n]
August 20, 2025 at 6:54 PM
Happy to share that our paper has been accepted to #EMNLP2025 main conference!
🚨 New Preprint! 🚨
Smarter LLMs are more selfish.
We show reasoning-enhanced models significantly prefer greed over cooperation. The more LLMs reason, the worse they cooperate.
👇 THREAD 👇
[Link to paper: arxiv.org/abs/2502.177...
[1/n]
Smarter LLMs are more selfish.
We show reasoning-enhanced models significantly prefer greed over cooperation. The more LLMs reason, the worse they cooperate.
👇 THREAD 👇
[Link to paper: arxiv.org/abs/2502.177...
[1/n]
May 22, 2025 at 5:27 PM
🚨 New Preprint! 🚨
Smarter LLMs are more selfish.
We show reasoning-enhanced models significantly prefer greed over cooperation. The more LLMs reason, the worse they cooperate.
👇 THREAD 👇
[Link to paper: arxiv.org/abs/2502.177...
[1/n]
Smarter LLMs are more selfish.
We show reasoning-enhanced models significantly prefer greed over cooperation. The more LLMs reason, the worse they cooperate.
👇 THREAD 👇
[Link to paper: arxiv.org/abs/2502.177...
[1/n]
🚨 New Paper! 🚨
“Groups that communicate diverse information make wiser decisions.” Right? Not always—for LLMs, just like humans.
We bring this scrutiny to AI by introducing the Hidden Profile paradigm to assess how multi-agent LLMs actually reason together.
arxiv.org/abs/2505.115...
[1/n]
“Groups that communicate diverse information make wiser decisions.” Right? Not always—for LLMs, just like humans.
We bring this scrutiny to AI by introducing the Hidden Profile paradigm to assess how multi-agent LLMs actually reason together.
arxiv.org/abs/2505.115...
[1/n]
May 20, 2025 at 5:18 PM
🚨 New Paper! 🚨
“Groups that communicate diverse information make wiser decisions.” Right? Not always—for LLMs, just like humans.
We bring this scrutiny to AI by introducing the Hidden Profile paradigm to assess how multi-agent LLMs actually reason together.
arxiv.org/abs/2505.115...
[1/n]
“Groups that communicate diverse information make wiser decisions.” Right? Not always—for LLMs, just like humans.
We bring this scrutiny to AI by introducing the Hidden Profile paradigm to assess how multi-agent LLMs actually reason together.
arxiv.org/abs/2505.115...
[1/n]
Reposted by Yuxuan Li
😬In sims, 1/100 Asian-coded agents will join a protest if told they shouldn't. All 100 Black-coded agents will join anyway.
New #Facct2025 paper led by @yuxuanli1225.bsky.social
Actions Speak Louder than Words: Agent Decisions Reveal Implicit Biases in Language Models
sauvikdas.com/papers/64/se...
New #Facct2025 paper led by @yuxuanli1225.bsky.social
Actions Speak Louder than Words: Agent Decisions Reveal Implicit Biases in Language Models
sauvikdas.com/papers/64/se...
May 13, 2025 at 2:26 PM
😬In sims, 1/100 Asian-coded agents will join a protest if told they shouldn't. All 100 Black-coded agents will join anyway.
New #Facct2025 paper led by @yuxuanli1225.bsky.social
Actions Speak Louder than Words: Agent Decisions Reveal Implicit Biases in Language Models
sauvikdas.com/papers/64/se...
New #Facct2025 paper led by @yuxuanli1225.bsky.social
Actions Speak Louder than Words: Agent Decisions Reveal Implicit Biases in Language Models
sauvikdas.com/papers/64/se...
Reposted by Yuxuan Li
We just found out that @yuxuanli1225.bsky.social 's paper was accepted to #FAccT2025
A true testament to his hard work and determination!
A true testament to his hard work and determination!
In this new pre-print, my student @yuxuanli1225.bsky.social outlines how language agents making shockingly biased decisions, even when their words seem "unbiased". Also, the latest models are better at hiding bias, but it still drives what they do.
It's also his FIRST Ph.D. paper!
Please boost :)
It's also his FIRST Ph.D. paper!
Please boost :)
🚨 Advanced LLMs may sound unbiased—but it's a ruse. We show that demographically-informed language agents reveal stark, implicit biases in decision-making.
New preprint: “Actions Speak Louder Than Words: Agent Decisions Reveal Implicit Biases in Language Models”
👉 arxiv.org/abs/2501.17420
[1/n]
New preprint: “Actions Speak Louder Than Words: Agent Decisions Reveal Implicit Biases in Language Models”
👉 arxiv.org/abs/2501.17420
[1/n]
April 11, 2025 at 7:56 PM
We just found out that @yuxuanli1225.bsky.social 's paper was accepted to #FAccT2025
A true testament to his hard work and determination!
A true testament to his hard work and determination!
Reposted by Yuxuan Li
How does using GenAI tools reshape knowledge workers’ critical thinking? Our #CHI2025 paper studied 319 knowledge workers to dive into this question. w/@advaitsarkar.bsky.social @levlevlev.bsky.social Ian, Sean, Richard, Nick @msftresearch.bsky.social @hcii.cmu.edu
www.microsoft.com/en-us/resear...
www.microsoft.com/en-us/resear...
March 28, 2025 at 8:38 PM
How does using GenAI tools reshape knowledge workers’ critical thinking? Our #CHI2025 paper studied 319 knowledge workers to dive into this question. w/@advaitsarkar.bsky.social @levlevlev.bsky.social Ian, Sean, Richard, Nick @msftresearch.bsky.social @hcii.cmu.edu
www.microsoft.com/en-us/resear...
www.microsoft.com/en-us/resear...
Reposted by Yuxuan Li
🔒 Our new #USEC2025 paper shows that by automating permission decisions based on prior decisions, we risk NORMALIZING privacy discomfort rather than reducing it.
"Modeling End-User Affective Discomfort With Mobile App Permissions"
Paper: sauvikdas.com/papers/60/se...
"Modeling End-User Affective Discomfort With Mobile App Permissions"
Paper: sauvikdas.com/papers/60/se...
February 26, 2025 at 2:22 PM
🔒 Our new #USEC2025 paper shows that by automating permission decisions based on prior decisions, we risk NORMALIZING privacy discomfort rather than reducing it.
"Modeling End-User Affective Discomfort With Mobile App Permissions"
Paper: sauvikdas.com/papers/60/se...
"Modeling End-User Affective Discomfort With Mobile App Permissions"
Paper: sauvikdas.com/papers/60/se...
Reposted by Yuxuan Li
🚨 In our new #USEC2025 paper, we introduce a secure, physically-intuitive RFID tag that users actually trust.
On-demand RFID: Improving Privacy, Security, and User Trust in RFID Activation through Physically-Intuitive Design
👉 Paper: sauvikdas.com/papers/62/se...
#Privacy #AcademicSky #Security
On-demand RFID: Improving Privacy, Security, and User Trust in RFID Activation through Physically-Intuitive Design
👉 Paper: sauvikdas.com/papers/62/se...
#Privacy #AcademicSky #Security
February 24, 2025 at 2:54 PM
🚨 In our new #USEC2025 paper, we introduce a secure, physically-intuitive RFID tag that users actually trust.
On-demand RFID: Improving Privacy, Security, and User Trust in RFID Activation through Physically-Intuitive Design
👉 Paper: sauvikdas.com/papers/62/se...
#Privacy #AcademicSky #Security
On-demand RFID: Improving Privacy, Security, and User Trust in RFID Activation through Physically-Intuitive Design
👉 Paper: sauvikdas.com/papers/62/se...
#Privacy #AcademicSky #Security
Reposted by Yuxuan Li
🚀 Our paper has been accepted to #CHI2025! 🎉
arxiv.org/abs/2409.12000
As AI technologists and data workers increasingly enter the news industry, cross-functional collaboration with journalists is becoming essential. But how do these collaborations unfold in practice?
arxiv.org/abs/2409.12000
As AI technologists and data workers increasingly enter the news industry, cross-functional collaboration with journalists is becoming essential. But how do these collaborations unfold in practice?
"It Might be Technically Impressive, But It's Practically Useless to us": Motivations, Practices, Challenges, and Opportunities for Cross-Functional Collaboration around AI within the News Industry
Recently, an increasing number of news organizations have integrated artificial intelligence (AI) into their workflows, leading to a further influx of AI technologists and data workers into the news i...
arxiv.org
February 14, 2025 at 4:26 PM
🚀 Our paper has been accepted to #CHI2025! 🎉
arxiv.org/abs/2409.12000
As AI technologists and data workers increasingly enter the news industry, cross-functional collaboration with journalists is becoming essential. But how do these collaborations unfold in practice?
arxiv.org/abs/2409.12000
As AI technologists and data workers increasingly enter the news industry, cross-functional collaboration with journalists is becoming essential. But how do these collaborations unfold in practice?
Reposted by Yuxuan Li
@hankhplee.bsky.social's MSR internship work is getting quite a bit of attention!
Following a year where he was recognized with both a CHI Best Paper and a USENIX Security Distinguished Paper, it looks like I will one day be best known for being Hank's advisor :p
Following a year where he was recognized with both a CHI Best Paper and a USENIX Security Distinguished Paper, it looks like I will one day be best known for being Hank's advisor :p
Microsoft's own research confirms something that was already pretty obvious: relying on a text generating machine to come up with answers erodes critical thinking, and is a method favoured by those who never liked doing critical thinking in the first place
advait.org/files/lee_20...
advait.org/files/lee_20...
advait.org
February 9, 2025 at 8:37 PM
@hankhplee.bsky.social's MSR internship work is getting quite a bit of attention!
Following a year where he was recognized with both a CHI Best Paper and a USENIX Security Distinguished Paper, it looks like I will one day be best known for being Hank's advisor :p
Following a year where he was recognized with both a CHI Best Paper and a USENIX Security Distinguished Paper, it looks like I will one day be best known for being Hank's advisor :p
Reposted by Yuxuan Li
🚨Trapped by infinite scroll & autoplay on social media? Our TOCHI paper introduces a system that reduces distraction by 4x by suppressing these and other dark patterns.
Purpose Mode: Reducing Distraction through Toggling ACDPs on Social Media Web Sites
📝 hankhplee.com/papers/purpo...
Purpose Mode: Reducing Distraction through Toggling ACDPs on Social Media Web Sites
📝 hankhplee.com/papers/purpo...
February 7, 2025 at 3:03 PM
🚨Trapped by infinite scroll & autoplay on social media? Our TOCHI paper introduces a system that reduces distraction by 4x by suppressing these and other dark patterns.
Purpose Mode: Reducing Distraction through Toggling ACDPs on Social Media Web Sites
📝 hankhplee.com/papers/purpo...
Purpose Mode: Reducing Distraction through Toggling ACDPs on Social Media Web Sites
📝 hankhplee.com/papers/purpo...