Two co-authored papers from Tohoku NLP Group (migrated to @tohoku_nlp) were accepted by #eacl2023 (1 main, 1 findings)! Both are about the symbolic reasoning ability of neural learners. AACL SRW 2022 had also encouraged our non-archival Findings work with the best paper award. Thanks! x.gd/aaclsrw2022

btw, I'm still more than a year away from it, but I'll probably be looking for a #postdocposition somewhere in the US, so hit me up if you'd like to chat! Commonsense reasoning in/with large neural networks is my jam (warning: might talk your ear off)

a few years ago we had BERTology, will we now have ChatGPTology?

I'm at #eacl2023 in Dubrovnik! Exhausted as hell from jet lag and crowds but come talk to me anyway 😂

I wonder if someday we'll start having careers like "AI teacher", "AI doctor", ... where your job is to tailor finetuning curricula, run diagnostics, etc

Peeking out of twitter hibernation just to give a big thanks to everyone at MBZUAI who gave me a very warm welcome during my four-week stay in Abu Dhabi ❤️ Hope to see you guys again!

I officially have my own budget(s!) for the first time, and it feels like the world changed. All the conferences! All the GPUs! Invite friends to guest speak! Visit friends' labs! 💸

I'm at #ACL2024, feel free to say hello! I'll also have a poster at the NLRSE workshop. It's an older paper but I'm excited to revisit it in more recent times & hear opinions 🙌

I'll be attending Conference on Language Modeling next week to present ACORN! Very excited to see how the first iteration will go. (Also looking forward to crying over the legendary Philly sandwiches 🤩)

My poster is Session 2, #25. Hope to see you there! (but I cannot promise coherence, 13h jet lag is *brutal* 🥲)

This study reveals when LLMs actually figure out their answers - before or during explanations. This research investigates how LLMs internally determine their answers during Chain-of-Thought reasoning, revealing whether they follow a "think-to-talk" (predetermined conclusion) or