Computer science people have developed an unintentional form of gatekeeping by virtue of being extremely bad at explaining/naming things, which results in simple concepts sounding far more complex than they are.
There is a definite consequence of this rule: international students will pursue less risky research questions. Hard questions and hard projects take longer to finish.
The overall quality of PhD research will drop, in social sciences, in engineering, in health… in everything!
PhDs in the social sciences generally take 5-6 years
what this means is that you may be admitted to do a PhD, but then have your visa extension application denied between years 4 and 5
This is such a disaster. A CS PhD takes on average 5–6 years. So, my international students will fall out of status before finishing their research program. Disaster.
Delighted to announce that I will be presenting our work “𝐁𝐚𝐲𝐞𝐬𝐢𝐚𝐧 𝐒𝐨𝐜𝐢𝐚𝐥 𝐃𝐞𝐝𝐮𝐜𝐭𝐢𝐨𝐧 𝐰𝐢𝐭𝐡 𝐆𝐫𝐚𝐩𝐡-𝐈𝐧𝐟𝐨𝐫𝐦𝐞𝐝 𝐋𝐚𝐧𝐠𝐮𝐚𝐠𝐞 𝐌𝐨𝐝𝐞𝐥𝐬” at 𝑨𝑪𝑳 2026. Meet me and @gergerliguven at poster session A, Sunday July 5th from 11:00 to 12:30 .
1/5
Think you're good at social deduction games? 🚀 Try your hand against our newest game-playing agent, GRAIL, tomorrow at #ACL2026 🌴
@RRShahab will be presenting our work, Bayesian Social Deduction with Graph-Informed Language Models, at the poster session at 11am! 🧵👇
All code, models, and a new Avalon dataset are open-sourced. Check it out: https://t.co/eqoOxunatH
And if you are in San Diego for ACL 2026, let's meet and chat.
5/5
✅ The factor graph feeds LLM with beliefs, while the LLM provides priors to the graph
✅ GRAIL emits 10× fewer output tokens than chain-of-thought reasoning agents, making it efficient and fast against humans
3/5
✅ GRAIL punches above its weight: smaller GRAIL agents achieve winrates comparable to larger DeepSeek reasoning agents
✅ Against humans, GRAIL notched a 67 % win rate and was rated more helpful than both humans and AI baselines
4/5
In this paper we introduce GRAIL, a hybrid language agent that tackles the social deduction game Avalon
✅ GRAIL uses belief propagation message passing over a factor graph to estimate the probability of each player's role
2/5
Delighted to announce that I will be presenting our work “𝐁𝐚𝐲𝐞𝐬𝐢𝐚𝐧 𝐒𝐨𝐜𝐢𝐚𝐥 𝐃𝐞𝐝𝐮𝐜𝐭𝐢𝐨𝐧 𝐰𝐢𝐭𝐡 𝐆𝐫𝐚𝐩𝐡-𝐈𝐧𝐟𝐨𝐫𝐦𝐞𝐝 𝐋𝐚𝐧𝐠𝐮𝐚𝐠𝐞 𝐌𝐨𝐝𝐞𝐥𝐬” at 𝑨𝑪𝑳 2026. Meet me and @gergerliguven at poster session A, Sunday July 5th from 11:00 to 12:30 .
1/5
Brilliant idea! Next up: Apple randomly reboots your Mac if you're building competing tech, Gmail silently edits your email if you mention rival platforms, and Tesla Autopilot swerves if it detects you're working on self-driving cars.
All in the name of safety, of course. Because malicious actors controlling the world’s operating systems, inboxes and cars would be extremely dangerous!
The Midwest Machine Learning Symposium (MMLS) 2026 will happen at Purdue University!
📍 West Lafayette, IN
📅 June 24–25, 2026
🔗 https://t.co/LXuUUbWWSP
📌 Poster submission deadline: May 24
We have an amazing lineup of plenary speakers: Tong Zhang, Jennifer Neville @ProfJenNeville, Mohit Bansal @mohitban47, Joyce Chai.
Looking forward to seeing you there!
@PurdueCS@PurdueECE@PurdueStats
Looking forward to giving a keynote at the Midwest Machine Learning Symposium (MMLS) 2026 (being held at Purdue University this year) & meeting folks from all the strong universities in the midwest, with their inspiring, long tradition of these exciting symposiums! 🙂
👇👇