Glenn Matlin
@GlennMatlin
PhD Computer Science (Intelligent Systems) ๐ง ๐ค @gtcomputing ๐๐ป@matsprogram Fellow ๐ฅ๐
๊ฐ์
November 2021
258 ํ๋ก์ ์ค 774 ํฌ
๐คWhere in trillions of pre-training tokens do capabilities actually come from?
๐ฅIn our new COLM 2026 paper, we trace the capability provenance of OLMo3-7B across Dolma3 using gradient-based training data attribution
๐งต
Paper:
Code:
Explorer:
๋ ๋ณด๊ธฐ
