Seoul National University Releases MineAmongUs: VLM Agents Learn to Lie in Embodied Social Settings
SeoulNatlUniv · hf · 2026-09-01
A Seoul National University team released the paper Lies We Can See: Joint Verbal and Non-Verbal Deception by VLM Agents in Embodied Social Interactions.
- MineAmongUs: a 3D multimodal Among Us environment for studying deception by embodied VLM agents across both verbal and non-verbal action channels.
- The ARIA harness systematically measures joint deception capabilities.
- Key finding: non-verbal channels are critical to winning deception games; verbal-only deception is far less effective.
More from Research
- Evolution Strategies Boost LLM Reasoning Coverage — yeewhye · 2026-09-01
- New Paper: Learning Agile Perceptive Traversal of Sparse 3D Structures for Humanoids — ChongZzZhang · 2026-09-01
- Talk: Eliciting Any-Order Inference from Any-Order Models — NandoDF · 2026-09-01
- Why Diffusion is Slow and How Flow Matching Speeds It Up — NandoDF · 2026-09-01
- LLM Engineering Lacks Discipline: Moving from Intuition to Rigor — camerongreen95 · 2026-09-01
- Gobanorobotics releases Toutatis v1: trains 99%+ reliable robot controllers from few demos — lukas_m_ziegler · 2026-09-01