🔥The 9th Round of Easy Loan, Earn $40 Reward is in progress❗️ ⏰ Promotion Period: January 15th - Feburary 15th, 2025 👉 Register now and check more details at gate.io/campaigns/358

New paper with Roger Levy: Prompt-based methods may underestimate large language models' linguistic generalizations. Preprint: lingbuzz.net/lingbuzz/007313 🧵👇

Let X = exp(U²/2) and Y=exp(V²/2) where U,V are joint normal with mean 0, variance 1, and 50% correlation. Then, E[X|Y] = 2Y E[Y|X] = 2X They are each, on average, twice as big as the other!

Thanks to several hundred of you for doing some crosswords for science! We’re excited to start some analysis soon but if you haven’t yet, please enjoy some puzzles: iliad.stanford.edu/experiments/te… And thanks again to boswords.org!

How can we leverage a theory of linguistic meaning to build machines that think in more human-like ways? New paper with Lio Wong, Alex Lew, noahdgoodman, Vikash Mansinghka, Jacob Andreas, and Josh Tenenbaum. arxiv.org/abs/2306.12672

🚨🚨New Paper Announcement (to appear in TACL) 📜 from me, Tiago Pimentel, Clara Isabel Meister, @ryandcotterell and Roger Levy: Testing the Predictions of Surprisal Theory in 11 Languages arxiv.org/abs/2307.03667 🌎

What is next in this sequence: 1. What's happening? 2. What is happening?! 3. ...

lazy-eyed susans

👋I boost a lot of job opportunities on here and now it's time to boost my own! I'll be arriving at #Stanford in fall of 2024 and I'm looking for awesome people to help me figure out language. 🧵👇

no, that’s not the kind of latex support i want

Have you ever done a dense grid search over neural network hyperparameters? Like a *really dense* grid search? It looks like this (!!). Blueish colors correspond to hyperparameters for which training converges, redish colors to hyperparameters for which training diverges.

Now out in PNASNews! Large-scale reading evidence that next-word predictability effects in humans are driven by *inference* (logarithmic in predictability) rather than preactivation (linear in predictability). pnas.org/doi/10.1073/pn… 1/10

🔔🌟 New Preprint Alert 🔔🌟 “An Information-Theoretic Analysis of Targeted Regressions during Reading” with Tiago Pimentel , Clara Isabel Meister , @ryandcotterell - Psycholinguistics 🧠 Computational Modeling 🤖 Crosslinguistic Studies 🌍 Information Theory 📡 osf.io/preprints/psya…

Need a reminder about how to report effect size or confidence intervals for your frequentist stats model? This looks very promising resource: matthewbjane.quarto.pub/guide-to-effec…

Wow that’s a really cool visualization of an LLM completion distribution. I want to see more interactive visualizations like this.

It was so much fun to be a part of this production.

It was an amazing experience to film our music video at the beautiful historic Estate of Mind Oakhurst Manor back in November. Very sadly, I learned that they suffered a devastating fire there on March 1st. Thankfully and most importantly no one was injured. But they are - (1/2)

Now that the eclipse is within the short range, global ensembles add less value while more high resolution models become available. I switched my eclipse page to use the High Resolution Ensemble Forecast (HREF), including a cloud layer breakdown: arctic.som.ou.edu/tburg/products…

Once, we ran a study on Prolific and a participant wrote on Reddit that the study “Felt like I was losing the will to live.” I went on the Prolific Subreddit (24k members!) and asked what matters. Here is what they told me. A thread on happier participants and better studies 1/9