Can Frontier Models See Past a Decoy? Elorian Tests Visual Reasoning
benchmarksLatent SpaceOct 29, 2026
A new decoy counting benchmark exposes a gap between seeing and reasoning. Elorian’s Glimpse tries to bridge it by thinking with images.
Impact: LowConfidence: LowNovelty: Incremental
benchmarkslatent-space