Anya Ji

Longer chain-of-thought = slower inference, more context rot, and ballooning compute. So what if the model could decide for itself when to go parallel? Our new BAIR blog breaks down Adaptive Parallel Reasoning (APR) — the next paradigm in inference-time scaling. 🧵

@jiayi_pirate Wish you all the best Jiayi!!

Today is my last day at xAI. It's been an intense, memorable year building in the frontier — I'm grateful to have worked with and learned from such talented and supportive colleagues. The journey itself is the reward. So long, and thanks for all the fish! 🐬

A couple years (!) in the making: we’re releasing a new corpus of embodied, collaborative problem solving dialogues. We paid 36 people to play Portal 2’s co-op mode and collected their speech + game recordings Paper: arxiv.org/abs/2512.03381 Website: berkeley-nlp.github.io/portal-dialogu…

Can Transformers Do Everything, and Undo It Too? Check out my blog on whether language models are surjective, injective, or invertible! astro-eric.github.io/blogs/surjecti…

Happy to announce the first workshop on Pragmatic Reasoning in Language Models — PragLM @ COLM 2025! 🧠🎉 How do LLMs engage in pragmatic reasoning, and what core pragmatic capacities remain beyond their reach? 🌐 sites.google.com/berkeley.edu/p… 📅 Submit by June 23rd

📝 Submit 4–8 page papers (non-archival, work under review welcome!) 📅 Deadline: June 23rd AoE 🔗 OpenReview: openreview.net/group?id=colmw… Let's advance our understanding of pragmatic competence—in both machines and minds!

@hawkrobe Did feel surreal! :)

Congratulations to @__anyaj , @noriyuki_kojima , Noah Rush , @alsuhr , @wkvong , @hawkrobe and @yoavartzi for winning the EMNLP 2022 Best Long Paper award.