BASIS uses balanced hashing and invariant scalars to sketch activations, cutting memory to O(L*R*N) while matching exact backprop performance on GPT training at R=32.
Evolution strategies as a scalable alternative to reinforcement learning
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
BASIS: Balanced Activation Sketching with Invariant Scalars for "Ghost Backpropagation"
BASIS uses balanced hashing and invariant scalars to sketch activations, cutting memory to O(L*R*N) while matching exact backprop performance on GPT training at R=32.