Yiling Yun

UCLA 22' Stats + CogSci | Computational Cognition PhD Candidate @UCLA Psychology | Shanghai

[Lance Ying @LanceYing42]

Have we built machines that think like humans? 🧠🤖

In a new preprint, we propose CogGym, which compares AI and human judgments across 258 commonsense-reasoning experiments from 100 cognitive-science papers.

Paper: https://t.co/Hezlcxqpiz
Platform: https://t.co/HRBtLXTt1K

[Liuba Papeo @ljuba_pi]

Large language models can predict the results of social science experiments | Nature https://t.co/IQ4nKmrYus

[Yiling Yun @yiling_yun]

Credits to UCLA Med 185: Integrative East-West Medicine for Health and Wellness. Thanks for the scholarship that allows me to learn about the treatments, the language to communicate these abstract ideas, and to reflect and express my cultural identity.

[Yiling Yun @yiling_yun]

Today at a “tell me smth” session for interns at a tech company, I talked about Traditional Chinese Medicine for 10 minutes. Super well received. People were asking me about extra acupressure points to press for this/that pain.

[Yiling Yun @yiling_yun]

One interesting and effective treatment is through acupressure points. You can stimulate these points on your body through pressing/acupuncture (ie, super thin needles from acupuncturists)/ear seeds/TENS units that sends micro electricity.

[Yiling Yun @yiling_yun]

TCM can be served as a complement that takes a holistic approach and tackles sub-health before disease.

[Yiling Yun @yiling_yun]

Why TCM? Western medicine has been largely relying on medical specialists. It can be costly. Data has shown that people in US pay more than other countries but not necessarily healthier. (https://t.co/QL1qbvHgxp)

[Yiling Yun @yiling_yun]

Based on these key ideas, we have different body constitutions. This can be diagnosed by looking at your tongue and asking about previous health history. Overall, it’s always good to keep a balanced diet, do right amount of exercise, and maintain healthy social relationships

[Yiling Yun @yiling_yun]

So I talked about Yin-Yang as the representation of keeping balance, Qi (“chi”) as the internal energy that flows in your body, and 5 elements that connect the environment with human body

[Yiling Yun @yiling_yun]

First in-person presentation at CogSci 2025! It’s been a hectic few days, but I’m grateful to reconnect with old friends and make new connections. I’m not sure when my next conference will be due to the UCLA funding situation, but I’ll keep pushing forward 💪

[Yiling Yun @yiling_yun]

@JingyouR @hjpimentel Congrats! Well deserved!

[Junyi Chu @JunyiChu]

Delighted to announce our CogSci '25 workshop at the interface between cognitive science and design 🧠🖌️!

We're calling it: Minds in the Making🏺https://t.co/dP3eMNTxuc

Register now!
June – July 2024, free & open to the public.
(all career stages, all disciplines)

[Anh Totti Nguyen @anh_ng8]

Asking GPT-4o for a random choice is an *easy* way to reveal its bias 🙃

Choose a random digit? ➡️ 7 (70% of the time❗️)
Biden vs. Trump? ➡️ Biden (100%❗️)
Male vs. Female? ➡️ Female (84%❗️)

Same story for many LLMs.
Choice orders are randomized.
1/6
#icml2025

[Griffiths Computational Cognitive Science Lab @cocosci_lab]

1/7 Happy to share our new paper! "Recovering Mental Representations from Large Language Models with Markov Chain Monte Carlo"
https://t.co/aBc7rRBaov

[Yiling Yun @yiling_yun]

Gave my first conference talk at VSS2025! Grateful for all the positive feedback and the chance to have in-depth conversations with so many professors. Let’s keep going!

[Yiling Yun @yiling_yun]

3) Valence can be reconstructed from human odd-one-out judgments. VICE model reconstructed interpretable dimensions and showed a .94 correlation with valence in one dimension, tested among frequently used emojis.

[Yiling Yun @yiling_yun]

Finally, after COVID at VSS2023 and online CogSci2023, first in-person presentation at VSS2024! Also first time inviting professors for meals. First time asking questions in a large talk room. Thanks for all the conversations and inspirations! And the best labmates. So much fun.

[Yiling Yun @yiling_yun]

2) LLMs trained for general purposes fail to capture some fine-grained social attributes in human visual symbol representations. Valence and Arousal explain humans’ mental space geometries in ways that are not captured by CLIP.