Pinned
Replying to @MaartenSap
Also, as a personal rule to preserve my well-being, I don't keep #NLProc Twitter logged in on my phone, so DMing me here will result in delayed responses. Email is better 📧
For more info on well-being in academia:
Maarten Sap (he/him)
@MaartenSap
retiring X acct: find me @maartensap.bsky
Working on #NLProc for social good.
Currently at @LTIatCMU, previously at @UWNLP, @MSFTResearch, and @allen_ai. 🏳🌈
- was excited to cover my group's work at this workshop, focusing on failures of agents w.r.t. human-facing aspects, like social intelligence, theory of mind, user reliance, and agentic and user-facing safety! Papers in 🧵Replying to @wzenusOur first speaker @MaartenSap from CMU and AI2, opened #FAGEN2026 with a talk on social intelligence in AI agents, from structured social world models (S3AP) to uncertainty communication. A great start to the day😄, connecting how agents understand social situations with how
- Changing the rules for reviewing requirements without people's consent / by blantantly ignoring people's pre-stated paper limits is bonkers, and making the only path to opt out "just withdraw your submissions" is also crazy...📢 ARR May 2026 cycle update We have published a blog post about the increased review assignments in the May cycle, what happened, and what comes next. Please read the full post here: aclrollingreview.org/may26-cycle-le… #ARR #ACL #NLProc
- Very excited to unveil this work! For training and evaluating AI that can truly interact and collaborate with humans, we need to have better user simulators, and this work is a step towards that!Does LLM really need to be a helpful assistant all the time? No. If you want to simulate people, “perfectly helpful” could be the wrong objective. Meet OdysSim, a journey toward LLMs beyond assistants, as behavioral foundation models (10B tokens of real human behavior; 23 sim
- Presenting two papers! On Tuesday at 12:40pm (Room I; 12pm session): 1-2-3 Check: Enhancing Contextual Privacy in LLM via Multi-Agent Reasoning arxiv.org/abs/2508.07667 (followed by poster session at 1pm)
