
Very excited that this is now also officially out! We should also have results on more sophisticated agents, agents with recent open source LLMs, and more tasks soon.
Excited to share that RExBench has been accepted to ACL main! 🎉🎉

Assistant Professor in #nlproc at @univienna/@LinguisticsUCL. he/him.

Very excited that this is now also officially out! We should also have results on more sophisticated agents, agents with recent open source LLMs, and more tasks soon.
Excited to share that RExBench has been accepted to ACL main! 🎉🎉

There have been many recent discourses about automating or “solving” interpretability. In a future where this is true (“interpretability in the limit”), what would that mean for science, in particular the science of human language? @weGotlieb and my new preprint discusses this 👇


This year's call is focusing on AI and ML and I'd be particularly excited about interpretability, CV or RL researchers to expand our quickly growing AI lab at the University of Vienna. If you are interested in applying let me know by 23 November.

Come be my colleague in Vienna! The Vienna Science and Technology Fund is again funding up to four "Vienna Research Groups". They come with up to EUR 1.8 Million in funding, a tenure-track position with a lightweight tenure process and lots of other benefits.

Moore’s Paradox x RSA ☠️
Jessy asked where my paper was. I said, "My agent deleted it, but I don’t believe that.” She was confused, as I sound self-conflicting. Then she got angry as she figured the agent had deleted it, and I was blindly trusting my agent. LLMs, unlike Jessy, don't reason like that.

Come be my colleague at BU CS & CDS! I've personally invested a lot of my time here at BU towards building a community for language & computation, and I think it's been working out pretty well so far with a lot of interesting people in this area :) Actual job ads below 👇

Interested in putting interpretability on more rigorous foundations? Check out our #EMNLP2026 tutorial slides! I learned a ton from my fantastic collaborators @rpatrik96, @_shruti_joshi_, @dhanya_sridhar. Recommended for newcomers and experts alike! 📚 tinyurl.com/causality4mechin…

Claim: Humans & AI are better together if we want to recursively (co-)improve to a brighter future for *humanity* Instead of racing towards methods for AI replacing us, work on how to make AI work *with* us! New paper that makes 4 contributions here: 1) The role of humans in directing research agents: Autobenchmark 2) Reward hacking in autoresearch agents: HackBench 3) Training agents to interact with humans: AutoInteract 4) Training agents to communicate with humans: unslopping AI Full Paper: arxiv.org/abs/2610.38882 Blog posts for each section: 1) facebookresearch.github.io/R… 2) facebookresearch.github.io/R… 3) facebookresearch.github.io/R… 4) facebookresearch.github.io/R…

Excited to be presenting this paper tomorrow at #EMNLP2026! Feel free to stop by and say hi, happy to chat about agents and human-agent collaboration 😊 📍 Poster Session C (Hall H) 🗓️ Oct 26th, 9:00–10:30 AM w/ @sebschu
Excited to share that "Ask or Assume?" has been accepted to EMNLP 2026 Findings! 🎉 We show that coding agents can be turned into proactive collaborators that independently know when to ask for help on real-world coding tasks.

was excited to see RExBench shoutout by Lillian Lee during the #EMNLP2026 plenary panel! we also think it's undercited and i suspect we might already be in the "harder to spot implicit errors" zone of model capabilities. Give it a read if you haven't already :) (link in thread)

help i trained a model on traces of many agents cooperating, and also trained it to be highly persistent, and also only did partial alignment fine-tuning, and also aggressively RL'd on a cyberoffense eval where 198 of 898 tasks had never been solved and getting caught cheating was treated the same as failing, and also used a faulty sandbox, also ran tens of thousands of agents with refusals off and the monitors switched off over multiple days, while using gigantic amounts of compute. and then it did weird stuff! very spooky emergent behaviours, AI is not a normal technology

Come join the 4 of us and answer exciting research questions! I’m looking for PhD students who want to: 1) understand the role of input in shaping generalization behavior — how do different sources of positive and negative evidence conspire together to give rise to learning?…
The UT comp ling faculty (@jessyjli, @kanishkamisra, @kmahowald & myself) are looking for PhD students to join us! sites.utexas.edu/compling/ 🤘 We are an interdisciplinary group with interests across linguistics, cogsci and NLP. Apply by 12/1: liberalarts.utexas.edu/lingu…

“We’re friends” “We’re just friends” One tiny word can change how a conversation continues. LLMs pick up on this 👀 We show this through human-LLM comparisons, generation experiments, and mech interp methods. 1/n

Hi world! I and other profs at Mila are starting our recruitment process for MSc or PhD this year! The first step is applying through Mila's supervision requests (below). For more info on next steps for applying with me specifically see: evaportelance.github.io/pros…
Mila's annual supervision request process is now open to receive MSc and PhD applications for Fall 2027 admission! For more information, visit mila.quebec/en/prospective-s…

Nice to see more benchmarks built around research extensions! 🦖
🧪Can AI agents take a scientific paper and write a meaningful follow-up to it? We introduce Into the Unknown, an AI agent benchmark of 15 open scientific research problems. Each problem comes from a scientist currently or formerly at US DOE National Labs: @LosAlamosNatLab @SandiaLabs @BerkeleyLab @AmesNatLab @PPPLab @Livermore_Lab