Elorian and Scale AI have released Humanity's Sixth Sense (HSS), a benchmark designed to measure how well AI models reason about information a visual scene implies but does not explicitly show. The benchmark includes 522 questions across 288 images and 234 short video clips, covering temporal and causal dynamics, physical and spatial logic, social understanding, and abstract contextual inference.
Results show a wide gap between humans and machines. People achieve an aggregate pass@1 of 93.1, while the best AI system, GPT 6 Astra, reaches only 53.6. Chinese open-weight models fare worse, with GLM-5.3-flash at 26.8 and Kimi K3 at 25.5. The most frequent errors were missing a detail that determines the answer (21% of failures) and misidentifying an object, person, or role (20%).
The authors suggest the benchmark also tests whether world models emerge in large language models. The relatively strong performance of frontier LLMs on tasks that were thought to require explicit spatial world models counters the argument that such models are necessary for intuitive visual reasoning. The source, Import AI 476, reports these findings without additional commentary or conflicting data.