Paper Club #5 : Emmanuel Ameisen - Anthropic

​Join us on May 28th for the fifth edition of the Unaite Paper Club with Emmanuel, research scientist at Anthropic !

This talk will explore recent developments in language model interpretability: the science of understanding the internal structure of large language models like Claude.

We recommend getting some familiarity with the topic beforehand, by looking over some papers from Anthropic's interpretability team:

​- Case studies showing LLMs plan many tokens ahead of time, have universal representations across languages, and more (https://transformer-circuits.pub/2025/attribution-graphs/biology.html)

​- Models learn delicate mechanisms to precisely estimate text length (https://transformer-circuits.pub/2025/linebreaks/index.html)

Thursday, May 28th, 2026 - 18h30 @ Albert School - 18 Rue de Paradis, 75010 Paris

See you there !

Note: The number of spots is extremely limited. To keep the room genuinely engaged, we ask applicants to answer a few short questions on Luma about their current projects, research interests, or what they're hoping to take away from the session.

⚠️ Please make sure to cancel your registration on Luma if you can no longer attend. To keep our events fair for everyone, repetitive no-shows will lose priority access to future events.