Mind's Eye - Visual Conceptual Slippage: leaderboard

Mind's Eye (2026) is a multiple-choice benchmark of visuospatial reasoning tasks adapted from human intelligence tests, with procedurally generated items answered with chain-of-thought prompting. This board is accuracy in percent on an odd-one-out task among six panels; higher is better.

Metric: Accuracy (%) on the visual conceptual slippage (odd one out among six panels) task of Mind's Eye (6-option multiple choice, 100 procedurally generated items, chain-of-thought prompt, answers extracted by an LLM evaluator; keys uniformly distributed over the options); higher is better. Source: arxiv.org. Saturation forecast: Estimated already saturated. 18 models tracked. Updated 7 Oct 2026.

Top models

#ModelScoreOverall rank
1Gemini 2.5 Pro35.3#276
2PaliGemma-3B (MindsEye checkpoint unspecified)34.7
3llava-v1.6-mistral-7B30.6#3999
4Qwen 2.5 VL 32B Instruct30.2#1062
5GT-Thinker-7B25.7
6V-Thinker-8B (MindsEye checkpoint unspecified)25.5
7GPT-4o25.2#751
8Vision-G1-7B24.1
9O323.7#237
10MindsEye InternVL3-8B (checkpoint unspecified)23.7
11Phi-3.5-Vision Instruct (MindsEye checkpoint unspecified)23.5
12Llama 3.2 11B Vision23.1#4264
13Qwen 2.5 VL 7B Instruct22.7#1902
14SmolVLM-2.2B (MindsEye checkpoint unspecified)21.2
15Qwen 2.5 VL 3B Instruct21#2713

No result here: #1 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.

Interactive version: theaggregate.ai/benchmark?slug=mind-s-eye-visual-conceptual-slippage · How It Works · Data refreshed daily, snapshot 2026-11-21.