SlopCodeBench: Benchmarking How Coding Agents Degrade over Long-Horizon Tasks

3 points | by zdw 7 hours ago