BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//dlrtp-ops//EN
X-WR-CALNAME:Deep Learning RTP
BEGIN:VEVENT
UID:2026-07-15-deepseek-technical-report@dlrtp
DTSTAMP:20261116T201215Z
DTSTART:20260715T160000Z
DTEND:20260715T180000Z
SUMMARY:DLRTP Weekly — LLM Architecture in 2026 - DeepSeek Technical Repo
 rt
LOCATION:Frontier RTP\, 800 Park Offices Dr\, Durham\, NC
DESCRIPTION:LLM architecture in 2026\, read through the DeepSeek technical 
 report. Discussion-format — skim the paper if you can\, come anyway if y
 ou can't. Bring your lunch.\nPaper: https://arxiv.org/abs/2412.19437\nRSVP
 : https://www.meetup.com/deep-learning-rtp/
URL:https://www.meetup.com/deep-learning-rtp/
END:VEVENT
BEGIN:VEVENT
UID:2026-07-22-emergent-sparsity-frozen-random-cnn@dlrtp
DTSTAMP:20261116T201215Z
DTSTART:20260722T160000Z
DTEND:20260722T180000Z
SUMMARY:DLRTP Weekly — Emergent Sparsity in Frozen-Random-CNN Deep RL (Sc
 ott Norton)
LOCATION:Frontier RTP\, 800 Park Offices Dr\, Durham\, NC
DESCRIPTION:Playing Pong through 3 neurons: freeze a randomly initialized C
 NN\, train a deep RL agent on top\, and the agent spontaneously compresses
  all task-relevant information through as few as 1-3 of the 64 neurons in 
 its first fully-connected layer\, with no sparsity objective anywhere. Sco
 tt Norton presents his manuscript (arXiv pending) documenting the phenomen
 on in Atari Pong. No RL background required\; the setting is informal and 
 open to everyone. Find us in the Eagles room off the main coworking area.\
 nRSVP: https://www.meetup.com/deep-learning-rtp/events/315516897/
URL:https://www.meetup.com/deep-learning-rtp/events/315516897/
END:VEVENT
BEGIN:VEVENT
UID:2026-07-29-agents-lets-get-this-over-with@dlrtp
DTSTAMP:20261116T201215Z
DTSTART:20260729T160000Z
DTEND:20260729T180000Z
SUMMARY:DLRTP Weekly — Agents - Let's get this over with
LOCATION:Frontier RTP\, 800 Park Offices Dr\, Durham\, NC
DESCRIPTION:Paper review of Lilian Weng's "Harness Engineering for Self-Imp
 rovement". Informal and open to everyone\; suggestions for future papers a
 nd topics welcome.\nPaper: https://lilianweng.github.io/posts/2026-07-04-h
 arness/\nRSVP: https://www.meetup.com/deep-learning-rtp/events/315598882/
URL:https://www.meetup.com/deep-learning-rtp/events/315598882/
END:VEVENT
BEGIN:VEVENT
UID:2026-11-25-deep-learning-architectures-systems@dlrtp
DTSTAMP:20261116T201215Z
DTSTART:20261125T170000Z
DTEND:20261125T183000Z
SUMMARY:DLRTP Weekly — Deep Learning Architectures & Systems
LOCATION:Frontier RTP\, 800 Park Offices Dr\, Durham\, NC
DESCRIPTION:Let's continue with our survey of developments since GPT-3.\n\n
 Useful starting points:\n\n* [A Visual Tour of LLM of Modern LLM Architect
 ures](https://www.youtube.com/watch?v=CepbWmGie0E) (youtube 38min)\n* [A s
 lop summary of Reiner Pope's interview ](https://solmaz.io/dwarkesh-reiner
 -pope-llm-training-and-serving)on how LLMs are actually trained and served
 . (Has some useful diagrams)\n* [The Deepseek family of models](https://ww
 w.bentoml.com/blog/the-complete-guide-to-deepseek-models-from-v3-to-r1-and
 -beyond) are the most well documented of the lot for the last two years.\n
 \nInformal\, all levels welcome. Suggestions for future papers and topics 
 welcome.\nRSVP: https://www.meetup.com/deep-learning-rtp/events/317075504/
URL:https://www.meetup.com/deep-learning-rtp/events/317075504/
END:VEVENT
END:VCALENDAR
