AI RESEARCH
No Memorization, No Detection: Output Distribution-Based Contamination Detection in Small Language Models
arXiv CS.CL
•
ArXi:2603.03203v2 Announce Type: replace-cross CDD, or Contamination Detection via output Distribution, identifies data contamination by measuring the peakedness of a model's sampled outputs. We study the conditions under which this approach succeeds and fails on small language models ranging from 70M to 410M parameters. Using controlled contamination experiments on GSM8K, HumanEval, and MATH, we find that CDD's effectiveness depends critically on whether fine-tuning produces verbatim memorization.