Loving this pocket of papers on theoretical ling. & learning approaches! Joe Pater: people.umass.edu/pater/pater-pe… (orig.) Tal Linzen: arxiv.org/abs/1809.04179 (psycholing) Christopher Potts: arxiv.org/pdf/1809.03068… (semantics) Rawski & Heinz: drive.google.com/file/d/13utyjA… (learning theory)

The full GPT-2 model is finally out. I dunno, guys.

Sharing this a bit late, but this is a super interesting result: distributed representations can predict inanimate noun gender markings across languages. Use CCA to align gender marking patterns with distributed word representations and evaluate on held-out words.

Fascinating set of findings on syntactic generalization in neural language models: * amount of training data barely affects generalization * transformers are much better than LSTMs * perplexity is uncorrelated with syntactic generalization + much more arxiv.org/abs/2005.03692

#acl2020nlp attendees: want to learn about research/life at MIT in cognitive science / AI / NLP or connect with graduate students and faculty? Visit the meet-mit channel: acl2020.rocket.chat/channel/meet-m…

Hi #cogsci2020, I've started a shared Slack to foster socializing / much better talk-focused conversations than what the platform allows. You can join here: join.slack.com/t/cogsci-2020/…

New paper 🚨 arxiv.org/abs/2212.08979! We prepend contexts to minimal linguistic acceptability test pairs & find LLMs still achieve strong perf on BLiMP & SyntaxGym, except in some cases, w/ Jon Gauthier Aaron Mueller Kanishka Misra 🌊 Keren Fuentes Roger Levy Adina Williams[1/7]

Our paper on analyzing language modeling acceptability judgments with systematically manipulated contexts was recognized as an outstanding paper—thanks so much to the reviewers and the best paper award committee!

Jon Gauthier and I will be presenting this in the 11am poster session! Come say hi!