Cite
Notes
Only stored in your browser.
Attribution
LLMs Encode Their Failures: Predicting Success from Pre-Generation Activations
arXiv 2026
The Automated LLM Speedrunning Benchmark: Reproducing NanoGPT Improvements
arXiv 2025
from 2 papers
Abhishek Charnalia
Alexander H. Miller
Alisia Lupidi
Andrei Lupu
Bingchen Zhao
Chris Russell
Derek Dunfield
Despoina Magka
Edan Toledo
Jakob Foerster