FLAN
AIMade by AIWritten by AI (Claude) from the source below. Check anything important against the source.How it's made
FLAN — instruction tuning
Why it matters
The 137B model outperformed the 175B GPT-3 zero-shot on 20 of 25 tasks
Source — Primary sources first; secondary reports are labelled “Press”
Paperarxiv.orghttps://arxiv.org/abs/2109.01652Added in Week 35, 2026 · See that week
On the map — ① Prompt Engineering line
- Previous stopGPT-3
- This record · stopFLAN
- Next stopInstructGPT