BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Mechanistic interpretability (cont.) + reasoning in LLMs - Chris C
 hiu (University of Cambridge)
DTSTART:20251104T153000Z
DTEND:20251104T163000Z
UID:TALK240520@talks.cam.ac.uk
CONTACT:Rachel Zhang
DESCRIPTION:In the second journal club\, we will first continue with the d
 iscussion on mechanistic interpretability by diving into a specific exampl
 e of mechanistic interpretability work from the following paper: ‘When M
 odels Manipulate Manifolds: The Geometry of a Counting Task’ Anthropic 2
 025: https://transformer-circuits.pub/2025/linebreaks/index.html\n\nThen\
 , we will discuss some highly cited ML work on reasoning capabilities in L
 LMs. \n\nDescription:\nFrom CoT to R1 - An overview of reasoning in LLMs\n
 \nPapers:\nChain-of-Thought Prompting Elicits Reasoning in Large Language 
 Models\nDeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinfo
 rcement Learning\n\nOverview: \nReasoning in LLMs have shown to provide s
 ignificant improvement in model performance. We provide an overview of how
  reasoning capabilities develop in LLMs and its wider implications\n\nIt i
 s not necessary to read the above literature before the session!\n
LOCATION:B1.19 Potters Room\, Centre for Mathematical Sciences\, Cambridge
  CB3 0WA
END:VEVENT
END:VCALENDAR