Loading activity…
@jayden_teoh_
phd @mit_csail
AI
The reported MATH-500 score rose from 7.9% to 48.1% with the same sampling settings.
AI
Jayden Teoh’s preprint says the method removes a noisy gradient without adding parameters or measurable overhead.
AI
The OPAL team reports gains in selected language and image runs, but most experiments were conducted once.
X
