paper_type.json (339B)
1 { 2 "paper_type": "empirical", 3 "reason": "Runs experiments on frontier LLMs to quantitatively measure strategic dishonesty behavior, reports F1 scores for detection methods, and validates findings through steering experiments; the primary contribution is experimental findings about model behavior, not the benchmarks being evaluated." 4 }