A benchmark for evaluating AI model alignment and safety.
AI BenchmarkSafety TestAlignmentEvaluation
Current best complexity is 14:
1024-Check can be discovered in Concepts, categorized under AI Benchmark, Safety Test, Alignment, Evaluation. A benchmark for evaluating AI model alignment and safety.
It has been discovered by 4 boards and crafted 4 times, first appearing 6 months ago. 1024-Check can be created from 1 recipe, and goes on to appear as an ingredient in 2 others.
1024-Check was first discovered by Skitsboard 6 months ago. They also hold the lowest-complexity recipe, at 14.
Related concepts worth exploring include Syzygy, Blockworld, Feedback, Analysis, Test, Acid Test, and 6 more.