Collecting a high quality dataset with 4M unique phrases and 52M corresponding object masks helped SAM 3 achieve 2x the performance of baseline models. Kate, a researcher on SAM 3, explains how the data engine made this leap possible.

🔗 Read the SAM 3 research paper:… pic.twitter.com/gfCCBQUChl

— AI at Meta (@AIatMeta) 20 novembre 2025

Collecting a high quality dataset with 4M unique phrases and 52M corresponding object masks helped SAM 3 achieve 2x the performance of baseline models. Kate, a researcher on SAM 3, explains how the data engine made this leap possible. Read the SAM 3 research paper:

→ View original post on X — @aiatmeta,