Mike Lewis

@ml_perception

Llama3 pre-training lead. Partially to blame for things like the Cicero Diplomacy bot, BART, RoBERTa, attention sinks, kNN-LM, top-k sampling & Deal Or No Deal.

Vibe and topics

Based on 62 recent X posts · click a vibe for evidence

Vibe

Teaching

Teaching22.7%

Announcing21.5%

Informing20.2%

Hopeful15.8%

Supportive15.6%

Humorous2%

Provocative2%

Topics

LLM Scaling

LLM Scaling29.7%

Diplomacy Agent17%

BART Pre-training14%

Llama Models12.1%

MARGE Pre-training11.4%

Contrastive Decoding8.7%

MegaByte Architecture7.2%