Marco

/en/country/jp

📍 Fujikawaguchiko, Japan 🇯🇵 — Local Time 06:00:19 AM

YouTube video by Michael Saxon (NLP & Generative AI research) The Science of Benchmarking Panel (NeurIPS 2025 Tutorial)

I have bootlegged our panel discussion from our NeurIPS "Science of benchmarking" tutorial to youtube!

Featuring @ofirpress.bsky.social @saining.bsky.social @idavidrein.bsky.social @efleisig.bsky.social and Wenda Xu

www.youtube.com/watch?v=d_zX...

I still don't get how this would work. Is there an ACL typst style template?

I always thought you were forced to use LaTeX or word due to those style templates (and some downstream post processing constraints). Is that not the case?

I'd love to see typst get popular in NLP/ML.

Lil Language Models

Have you stayed in touch with the RC community since attending?

I see what you mean about the minibatch being too short, but to me it would be worth it just to get plugged in.

Mini retreats - Recurse Center Can’t take six or 12 weeks away from your life? Do a mini retreat and come to RC for one week.

I'd love it if they brought back the mini-retreat, but it's paused now :/

Even a 3 week one would be good. I want to be part of the RC community so bad, but 6 weeks is tough (and now that I am working full time, functionally impossible).

Check out Heart's Light next time you are in Shibuya!

maps.app.goo.gl/hsZLqnsURMEV...

a close up of an old man 's face with the words terrible yes but great written below him ALT: a close up of an old man 's face with the words terrible yes but great written below him

Your LaTeX wizardry was both great and very painful for me.

Ok, so I didn't want to dunk on you in public but since you responded to this post, about 80% of the errors came from one of your macro.tex files from our LREC paper lmao.

/en/country/jp

📍 Tokyo, Japan 🇯🇵 — Local Time 05:00:10 AM

Some of our Lapras are abstract.

The ABCD framework for feedback This is a short and sweet way to get feedback on your work in a specific, helpful way!

For day 7 of #blogvent, I wrote about one of my favorite ways to ask for feedback on something I'm making!

cassidoo.co/post/abcd-fe...

No more ~~tears~~ errors, only warnings now.

I can live with warnings.

I've already blasted through my Claude daily limit 😬

LaTeX error "Package soul Error: Reconstruction failed."

I'm in the depths of LaTeX hell and got an appropriate compilation error.

A good read for the AI hot takes era.

They're Made out of Meat by Terry Bisson.

www.mit.edu/people/dpoli...

Join us today from 4:30 to 7:30 PM @neuripsconf.bsky.social Hall C,D,E #1006 for our poster on SmoothDiff, a novel XAI method leveraging automatic differentiation.
🧵1/6

YouTube video by Matt Godbolt [AoCO 0/25] The Advent of Compiler Optimisations

I guess it's also a week until the #AoCO2025 Advent of Compiler Optimisation! xania.org/202511/adven...
youtu.be/j-BwR-Cw0Gk

Submit your JuliaCon 2026 Minisymposium now! The JuliaCon 2026 Minisymposium CfP closes on December 2nd 2025 20:00 (CET). We are looking for as many ideas and proposals for Minisymposia as possible, we really want to be a conference of minisympo...

The #JuliaCon Global 2026 closes on December 2nd 2025 20:00 (CET).

Submit your proposal now!

discourse.julialang.org/t/submit-you...

#julia #julialang #conference #cfp

As a kid we had a station wagon with rear-facing seats that we used to read in while driving in the mountains.

We didn't know how good we had it.

Rapid Chess • fruitdealer2002 vs matvilla fruitdealer2002 (1806) played matvilla (1809) in a rated Rapid (10+0) game of chess. fruitdealer2002 won by checkmate after 39 moves. Click to replay, analyse, and discuss the game!

lichess.org/RtWctdjm

I regretted 21. Rf3 as soon as I did it but otherwise I was happy.

Today I played what I thought was my best game ever.

Stockfish disagreed.

#chess

Home Office Romance - Wikipedia

Source is from 「テレワァク与太話」("Home Office Romance" is the English title).

It's really cute, I recommend it.

A comic where the girl (a grad student) reads a full manga series instead of doing her work. The girl says "Ahh, I feel satisfied after finishing all 74 volumes." And the guy asks "Are you actually a graduate student?"

I'm in the "thesis writing" stage of my PhD.

Here's some cute bounding box functions for 2D shapes. A longer list here: iquilezles.org/articles/bbo...

I will be presenting our paper about tokenizer inequities at the main conference on Dec 4th at 11am (Poster Session 3) bsky.app/profile/cath...

Flyer for The PokeAgent Challenge at NeurIPS 2025. Sunday, Dec 7, 8–10:45 AM PST, Mezzanine Room 15AB, San Diego Convention Center. Two tracks: Track 1 (Battling) features competitive Pokémon battle bots; Track 2 (Speedrunning) features long-horizon RPG gameplay. Tagline: "How do we close the gap between specialist RL models and generalist LLM agents?" Speakers: Seth Karten (Princeton), Aaron Traylor, Minmin Chen (Google DeepMind), Jake Grigsby (UT Austin), Stephanie Milani (NYU/Johns Hopkins), Kiran Vodrahalli (Google DeepMind), Fei Fang (CMU), Yuke Zhu (UT Austin), Chi Jin (Princeton). Sponsored by Google DeepMind.

How do we close the gap between specialist RL and generalist LLM agents?

We're benchmarking it in Pokémon. Join us at the PokeAgent Challenge competition workshop @ NeurIPS 2025.

📍 Dec 7, 8AM
🎮 Track 1: Competitive Pokémon (game-theoretic reasoning)
🗺️ Track 2: Speedrunning (long-horizon planning)

I'm recruiting my first group of students at TTIC! If you're interested, please apply by December 9th and mention my name in your application

I made a Chrome plugin that converts your typing speed to tokens/second (TPS) so you can compare your output to LLMs.

150 WPM = roughly 3.3 tokens/sec

(🔊 Sound on)