Antoine Bosselut
@abosselut.bsky.social
Helping machines make sense of the world. Asst Prof @icepfl.bsky.social; Before: @stanfordnlp.bsky.social @uwnlp.bsky.social AI2 #NLProc #AI
Website: https://atcbosselut.github.io/
Website: https://atcbosselut.github.io/
(NAACL) Luca and @debjit-paul.bsky.social's work looks at how we can guide LLMs to generate logically sound arguments. Introducing FIPO: a fallacy-informed framework that improves preference optimization to help LLMs avoid logical fallacies in argumentation.
February 25, 2025 at 9:18 AM
(NAACL) Luca and @debjit-paul.bsky.social's work looks at how we can guide LLMs to generate logically sound arguments. Introducing FIPO: a fallacy-informed framework that improves preference optimization to help LLMs avoid logical fallacies in argumentation.
(NAACL) @bkhmsi.bsky.social showed that neuroscience localizers uncover brain-like functional specializations in LLMs. Across 18 LLMs, he discovered neuron groups that mirror the brain’s language, theory of mind, and multiple demand networks! Work w/ @mschrimpf.bsky.social @gretatuckute.bsky.social
February 25, 2025 at 9:18 AM
(NAACL) @bkhmsi.bsky.social showed that neuroscience localizers uncover brain-like functional specializations in LLMs. Across 18 LLMs, he discovered neuron groups that mirror the brain’s language, theory of mind, and multiple demand networks! Work w/ @mschrimpf.bsky.social @gretatuckute.bsky.social
(NAACL) @smamooler.bsky.social will introduce PICLe: a new method for in-context named-entity detection (NED) using AI annotations. No more human labels, which are tricky to get in specialized domains, as PICLe outperforms FS learning with gold annotations! Work with @trackingskills.bsky.social
February 25, 2025 at 9:18 AM
(NAACL) @smamooler.bsky.social will introduce PICLe: a new method for in-context named-entity detection (NED) using AI annotations. No more human labels, which are tricky to get in specialized domains, as PICLe outperforms FS learning with gold annotations! Work with @trackingskills.bsky.social
At ICLR, @agromanou.bsky.social will present INCLUDE: our new benchmark that spans 44 languages! Our special sauce ? We collect data that prioritizes regional knowledge, constructing an AI evaluation that reflects the true contexts where languages are used. Check out how your model measures up!
February 25, 2025 at 9:18 AM
At ICLR, @agromanou.bsky.social will present INCLUDE: our new benchmark that spans 44 languages! Our special sauce ? We collect data that prioritizes regional knowledge, constructing an AI evaluation that reflects the true contexts where languages are used. Check out how your model measures up!
4/ 🧑🎓 And that’s exactly what we observe. When we group questions by course, and courses by program, GPT-4 can pass 83% to 100% of courses at a 50% scoring threshold with 91.7% of courses passed across programs.
December 4, 2024 at 2:53 PM
4/ 🧑🎓 And that’s exactly what we observe. When we group questions by course, and courses by program, GPT-4 can pass 83% to 100% of courses at a 50% scoring threshold with 91.7% of courses passed across programs.
2/ 🌍 We evaluated GPT-3.5 & GPT-4 using 8 prompting strategies on 5,579 English and French questions from 50 STEM university courses at EPFL from both exams and assignments (including many in @icepfl.bsky.social)
December 4, 2024 at 2:53 PM
2/ 🌍 We evaluated GPT-3.5 & GPT-4 using 8 prompting strategies on 5,579 English and French questions from 50 STEM university courses at EPFL from both exams and assignments (including many in @icepfl.bsky.social)
