Martin Tutek
@mtutek
Followers
419
Following
3K
Statuses
410
Postdoc @ Technion | previously postdoc @ UKP Lab, TU Darmstadt | PhD @ TakeLab, UniZG | Working on interpretability & safety of LLMs.
Republic of Croatia
Joined October 2015
I'm URGENTLY looking for three emergency reviewers for ARR December, two for the Interpretability track and one for Generation. Reviews should be submitted asap, in the next ~24 hours. Please share for visibility 🙏
1
3
3
Happening now in Jasmine, come check out the poster @emnlpmeeting
Code prompting will appear at EMNLP Main 🎉🎉, see you soon in Miami! This paper was really cool to work on as it was one of the rare cases where every experiment just works. The paper shows nice improvements on performance and efficiency on conditional reasoning tasks, but...
0
0
9
@yoavgo What is cool here is that, just by seeing an instance in code form triggers the model to use more capacity for memorization - no tuning required! Which other input forms could elicit relevant capabilities? Could prompting in style of Gordon Ramsay improve food recipe generation?
0
0
2
RT @NobelPrize: BREAKING NEWS The Royal Swedish Academy of Sciences has decided to award the 2024 #NobelPrize in Physics to John J. Hopfiel…
0
14K
0
RT @LHaausing: I’m currently looking for internship opportunities. If any position that works on test-time improvement, code generation, o…
0
2
0
@somakaditya As AC, you were able to access them for review/discussion reminders, this doesn't seem to work now (it did two days ago)
1
0
0
RT @jacobandreas: Some thoughts on how to think about "world models" in language models and beyond:
0
55
0
@dirk_hovy Very important topic! Flashbacks from 2010 era papers where results of complex (ensembles of) models were often reported with a single run
0
0
0
RT @barbara_plank: ⁉Asking Multiple-Choice Questions to LLMs is 🔥, but do extracted first-token probabilities match the text output? 🤔 ❌No…
0
7
0
🚨PSA🚨 @NeurIPSConf submission is 01:00 PM PDT, @ReviewAcl AOE deadlines have spoiled me a bit 🥲
0
0
3