From the source
From the source
Cognition, in partnership with Applied Compute, introduced SWE-Check, a reinforcement learning-trained bug detection model that matches frontier performance on in-distribution evaluations and makes progress on out-of-distribution evaluations, while being faster and cheaper.
From the source
We've partnered with Applied Compute to put this to the test by collaborating to RL-train a bug detection model. The result is SWE-check , which matches frontier performance on internal in-distribution evals (delta F1 to Opus 4.6 goes from 0.09 to 0) and makes meaningful progress on out-of-distribution evals (delta F1 to Opus 4.6 goes from 0.49 to 0.29).
cognition.com
