Evaluating Small Models: Multiple-Choice Tests Are Easy to Game
mervenoyann · x · 2026-07-14
AI researcher Merve Noyant points out that when evaluating large models, most multiple-choice question assessments (MCQA) actually make it easy for models to "cheat": even if a model gets stuck in a loop or hallucinates, the structure of the multiple-choice options helps it organize and anchor the correct answer. Consequently, she prefers researching small models and emphasizes the importance of conducting "vibe tests."
Related event: Multiple-Choice Evaluations May Overestimate LLMs(2 posts)→
More from Research
- UK AISI: under $80 of fine-tuning fully restores filtered bio capabilities in open-weight models — connoraxiotes · 2026-10-27
- RFD4-Proteina runs on an RTX 3060: protein design from $0.001 per design — AllThingsApx · 2026-10-27
- University of Freiburg's Generative Intelligence Lab recruiting funded PhDs via ELLIS, apply by Oct 31 — yuyinzhou_cs · 2026-10-27
- Nature Physics: reprogrammable refractive index enables arbitrary multimode wave control for ML — MarioKrenn6240 · 2026-10-27
- Wakaru open-sources ontologize: labeling datasets from model activations — amplifiedamp · 2026-10-27
- EMNLP Oral paper traces multilingual representations in LLMs with cross-layer transcoders — ZhijingJin · 2026-10-27