We identify label errors in the test sets of 10 of the most commonly-used computer vision, natural language, and audio datasets, and subsequently study the potential for these label errors to affect b...

Support open-source research and human-accessible neural network speedrunning benchmarks

Adds experimental support for additional context tokens to Fused NA. Any number of partial attention results can be reduced into a final one as if their contexts were merged, which is just the same...