Megan Wei

🎶 We can often tell if a song is fast or slow or if a note is high or low. Music generation models are trained on vast amounts of data, but do they pick up music theory concepts like we do? 🎶 Our #ISMIR2024 paper explores this question! arxiv.org/abs/2410.00872 @ISMIRConf

@saelyne_yang @imjuhokim Yes, for sure!

@MeganJWei @imjuhokim Thank you Megan! We should catch up!

Excited to share that I’ve officially graduated with my Ph.D. and joined CMU as a Postdoctoral Fellow! I’m forever grateful to @imjuhokim for the unwavering support, and I'm thrilled to start this next chapter working with Jeff Bigham. Hit me up if you’re around Pittsburgh!

@juankonta DMs now open!

@MeganJWei DMs closed!

Excited to attend my first NeurIPS! Feel free to reach out if you want to chat about audio/speech/music generation, multimodal LLMs, and interpretability! Also, happy to give hiking/beach recs for San Diego and other parts of SoCal ☀️

How do people flexibly integrate visual & textual information to draw mental inferences about agents we've never met? In a new paper led by @LanceYing42, we introduce a cognitive model that achieves this by synthesizing rational agent models on-the-fly--presented at EMNLP 2025!

@SeungHeon_Doh @juhan_nam @keunwoochoi Congrats SeungHeon!!

I have completed Ph.D. journey! The title of my doctoral dissertation is "Connecting Audio and Natural Language for Music Annotation and Retrieval." I would like to express my deepest gratitude to my advisor, Professor @juhan_nam , and (unofficial co-advisor), Dr. @keunwoochoi

Thanks for inviting me @musaic_ucsd! Feel free to check out our work here: arxiv.org/abs/2410.00872

Check out our code, listen to our audio samples, and experiment with our dataset here! Website: brown-palm.github.io/music-theory Code: github.com/brown-palm/syn… Dataset: huggingface.co/datasets/megan… Special thanks to my collaborators @mik3fr33man @chrisdonahuey @jesu9 🎉

Guess what? Music generation models DO encode music theory concepts! The extent to which they do varies by model size and layer. This invites future research directions on controllable music generation by modifying the internal representations!

🎶 We can often tell if a song is fast or slow or if a note is high or low. Music generation models are trained on vast amounts of data, but do they pick up music theory concepts like we do? 🎶 Our #ISMIR2024 paper explores this question! arxiv.org/abs/2410.00872 @ISMIRConf

Can LLMs perform social reasoning? In this paper with @xuanalogue @katie_m_collins @MeganJWei, Lionel Wang, Cedegao Zhang, @adrian_weller and Josh Tenenbaum, we propose a model using LLM as a semantic parser and inverse planning as an inference engine. arxiv.org/abs/2306.14325

@shobsund @DeepMind @MIT_CSAIL @phillip_isola Congratulations! Super excited for you!

Just wrapped up a wonderful summer at @DeepMind and starting my PhD at @MIT_CSAIL! Excited to work with @phillip_isola and the rest of the Embodied Intelligence group!

Can a machine tell if you are happy about this tweet or not? Learn about emotional AI from Dr. Rana el Kaliouby @kaliouby of @Affectiva and hear about her new book Girl Decoded! Interviewed by @mememeganme #womeninAI #emotionalAI #girldecoded #womeninTech youtu.be/rMlVqYs6q1E

And a HUGE personal thank you to @kaliouby for donating 300 free copies of #girldecoded to our organization! We can’t wait to read it! [Correction from above tweet: Rana’s interviewer was WAI co-president Megan Wei @memeganmeme]

INTERVIEW SERIES: Did you know AI can help fight poaching AND protect the Staten Island Ferry? Learn about Prof. Fei Fang @fangf07 of @SCSatCMU and her work on computational game theory and multi-agent reinforcement learning, interviewed by @memeganmeme! youtube.com/watch?v=-qQftu…

INTERVIEW SERIES: Scalable, nonparametric, and unsupervised- @ta_broderick's research emphasizes all 3! Check out her interview with @memeganmeme to learn about Prof. Tamara Broderick’s work @MIT_CSAIL in Bayesian inference, graphical models, & more! youtu.be/T3cOoxspe0k

INTERVIEW SERIES: What makes an image memorable? Will designers' jobs be automated by AI? Dr. Zoya Bylinskii @zoyathinks of @Adobe answers these questions and talks AI-driven design, perception, human memory, and more! Interviewed by @memeganmeme! youtu.be/v5HHEZJ7920

INTERVIEW SERIES: What is the future of responsible AI? How can we strengthen the AI ecosystem in Mexico & Latin America? Watch our interview with @claudiamdelpozo of #EonResilienceLab @CMinds_co to find out, interviewed by @memeganmeme! #WomeninAIEthics youtube.com/watch?v=Yme3Bt…