Cognition launched SWE-1.7 on Wednesday, the most capable model it has trained, and with it made a specific claim about reinforcement learning that the broader AI research community will be evaluating ...
Some results have been hidden because they may be inaccessible to you
Show inaccessible results