Keunwoo Choi

hi music people, i wrote a tutorial on large language models and music information retrieval. of course it's called.. LLMs <3 MIR 🥁 have fun! llms-heart-mir.github.io/tutorial

we really need a moderator at the ISMIR mailing list. it has been fine without it as we were smaller, but we has passed the tipping point. it will only get worse in terms of both the frequency of spams & negative impact of them. i’d be happy to volunteer.

Working with world’s most respected engineers & researchers on the world’s most challenging problems is tech’s a house closer to the beach in Amagansett. We don’t need made up goals like the finance industry because we actually work on real problems.

Recently music streaming platforms are being flooded with AI-generated music. Use our new Spot-if-AI Chrome Extension to find out if the music you’re listening to on the Spotify Web Player is AI-generated. It is free, open-sourced on GitHub, and runs locally on your browser (1/6)

The AI Action Plan is out. Immediate reactions in this thread:

please book your accommodation at #ISMIR2025 either i) directly through links at ismir2025.ismir.net/accommodation or ii) just DIY. gtravelhost.com is not affiliated with us, it is a well-known scam. google.com/search?q=gtrav…

i just read the Voxtral paper, and somehow i've looked into a lot of details about the audio part. keunwoochoi.github.io/post.html?id=2…

so, time to introduce the the scaling law of the scaling laws. the average usefulness of each scaling law decreases exponentially, and therefore..

>>> Qwen3-Coder is here! ✅ We’re releasing Qwen3-Coder-480B-A35B-Instruct, our most powerful open agentic code model to date. This 480B-parameter Mixture-of-Experts model (35B active) natively supports 256K context and scales to 1M context with extrapolation. It achieves…

criminally underrated talk from @devanshtandon_ Gemini powers YouTube's Large Recommender Model by **tokenizing every video on youtube** (SemanticID) - a vocabulary several OoMs larger than English, CONTINUOUSLY PRETRAINED every day. can reason across titles/descriptions, throws…

It's an honor to host the RecSys x LLMs track at @aiDotEngineer SF! In this talk, I discuss semantic IDs, LLM-based data augmentation, and foundation rankers. Slides: eugeneyan.com/speaking/aie-2… Talk: youtube.com/watch?v=2vlCqD… This is my talk's semantic ID: 2vlCqD6igVA. I wish knew…

yes, AI and all good and IMO but.. did you know that the loudness normalization is still an unsolved problem? check out @gaudiolab's solution! (disclaimer: i'm an advisor & was working there. (but LM1 is actually good!))

re-inventing (or improving) the Amino Acid representation by HNet: work in progress..

i mean it is still a stochastic parrot. we just figured how to trach or incentivize the parrot so well.