Siddhant Paliwal

Four passing runs out of 80 separate first from second on Specific's private-code benchmark

Specific Labs scores coding agents on licensed production codebases. The best setup clears 38.8%. The analysis covers ten tasks at eight runs each, so every published score is a count of passing rollouts out of 80.