
Tomer Porian
@tomerporian
M.Sc. @ Tel Aviv University, Computer Science | ML
ID: 974532044
27-11-2012 19:31:40
44 Tweet
113 Followers
348 Following



ืืจืื ืืื ืืฉืืช ืืื ืืกืืกืืช ืื ืคืฉืื ืื ืืืื ืืื. ืืืืื ืืืจื 1973 ืืชืืืกืจื ืืฉื ืื ืืืขืืืช ืฉืืื ืขื ืฉืืืชื ืืืชื. ืืืื ืืืจื 1982 ืฉืงืข ืืืืืืื ืขืืืง ืฉืืืจ ืืืชื. ืืืื ืืื ื ืฉืืื 1200 ืืฉืจืืืื ืืืฃ ืืื ืืืืฉืืช ืืืืืขืืช ืืื ืื ืืชืืจื, ืื ืืฆืืขืจ, ืื ืืืื. ืขืืื ืืืจืืืช, ืขืืื ืืืื ืืืช, ืขืืื ืืื. ืชืืื ืื ื ืืื


Do you have a "tell" when you are about to lie? We find that LLMs have โtellsโ in their internal representations which allow estimating how knowledgeable a model is about an entity ๐ฃ๐ฆ๐ง๐ฐ๐ณ๐ฆ it generates even a single token. Paper: arxiv.org/abs/2406.12673โฆ ๐งต Daniela Gottesman







When all of the sudden puzzle pieces fall into right places and predicting the unknown starts to work, those are rare beautiful moments I am grateful for in science. Made by rare minds of Marianna Tomer Porian mehdi cherti github.com/LAION-AI/scaliโฆ arxiv.org/abs/2506.04598