🔥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

We're (still) in an Evaluation crisis!

What does it mean to be fluent in AI? 🤔

The United States has had a tremendous advantage in science and technology because it has been the consensus gathering point: the best students worldwide want to study and work in the US because that is where the best students are studying and working. 1/

Can decades old ideas from #psychology help fix critical issues in modern LLM alignment? 🤔 We're tapping into #BoundedRationality & 'satisficing principles' to build an alternate way to align LLMs. Our new #ICML2025 paper 👇 🧵 arxiv.org/pdf/2505.23729

my off-the-cuff take, after playing around with Gemini 2.5 a bit, is that solving Olympiad-style math problems gives it the insight and serendipity to also make significant strides in (and occasionally solve) research-level math

We will incur irreparable harm to the US academic leadership if we continue the current policies.

🤝 Humans + AI = Better together? Our #ACL2025 tutorial offers an interdisciplinary overview of human-AI collaboration to explore its goals, evaluation, and societal impacts 🤖

Even an LLM is ill defined imo. Is GPT2 an LLM? What distinguished LLM from an LM in a verifiable manner?

A language model is a probability mass function* p(y|x) where x is any sequence of tokens that belong to alphabet A, and y can be any sequence of tokens that belong to alphabet A. * I.e. sum_{y \in A*} p(y|x) = 1 for all x \in A*.

A string may get 17 times less probability if tokenised as two symbols (e.g., ⟨he, llo⟩) than as one (e.g., ⟨hello⟩)—by an LM trained from scratch in each situation! Our #acl2025nlp paper proposes an observational method to estimate this causal effect! Longer thread soon!