Larger and more instructable language models become less reliable

Add to your list(s) Download to your calendar using vCal

• José Hernández-Orallo Valencian Research Institute for Artificial Intelligence, UPV Leverhulme Centre for the Future of Intelligence, ITH, Cambridge
• Wednesday 09 October 2024, 18:00-19:00
• Syndics Room, 17 Mill Lane.

If you have a question about this talk, please contact Alva Markelius.

At this year’s first CHIA Early Career Community Seminar, Jose Hernandez Orallo will introduce his recent study published in Nature, which reveals that, as LLM families scaled up and shaped up, their models didn’t become more reliable. This seminar talk introduces some methodological innovations from the perspective of AI evaluation.

This talk is part of the Centre for Human Inspired AI Early Career Committee series.

This talk is included in these lists:

• Syndics Room, 17 Mill Lane

Note that ex-directory lists are not shown.