🔥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

really basic naive q but how does distributional semantics handle function words ?? like i cannot imagine them having meaningful vector representations

transitive ‘shop’ ?? what is going on in the midwest

postering at #blackboxnlp this thursday in-person!! talking about memory effects in production and abstract "soft" ordering preferences, i'd love to chat about this or anything nlp/cognition/psycholinguistics if you'll also be there!

I am at #CCN2024 CogCompNeuro, will be presenting our work "Brain-Like Language Processing via a Shallow Untrained Multihead Attention Network" Poster C57 on Friday! Joint work w/ Greta Tuckute, Antoine Bosselut and Martin Schrimpf Looking forward to all the discussions! :D

Language Models learn a lot about language, much more than we expected, without much built-in structure. This matters for linguistics and opens up enormous opportunities. So should we just throw out linguistics? No! Quite the opposite: we need theory and structure.

new paper! 🫡 we introduce 🪓AxBench, a scalable benchmark that evaluates interpretability techniques on two axes: concept detection and model steering. we find that: 🥇prompting and finetuning are still best 🥈supervised interp methods are effective 😮SAEs lag behind

i'm at ICLR this week presenting TopoLM as an oral(!) reach out if you want to chat about anything humans + machines (cogsci, hci, fairness, safety, interp, etc.)

presenting our work on inducing brain-like topography in Transformers today 2x w/ Hannes Mehrer! come by to talk neuro, interp, etc. talk — Session 4C, 4:15pm [Garnet 216-218] poster — Session 3, 10am [Hall 2B #599]

you should be using interp for evals 🌟