Probing shows visual foundation models vary widely in persistent spatial state support
mmlab-ntu · hf · 2026-11-02
mmlab-ntu's study probes whether visual foundation models maintain persistent spatial state representations across four levels: spatial structure, object persistence, predictive spatial update, and actionability. Key finding: models vary widely, and high readout accuracy does not guarantee history-based state updates or reliable goal-directed progress — relevant for embodied agents that must build and maintain an internal world state from streaming experience.
Related event: Study finds no visual foundation model fully maintains spatial state(3 posts)→
More from Research
- NVIDIA's Kumo Is a Family of Four Tabular Foundation Models, Not Just One — pandeyparul · 2026-11-02
- Aaroth shares Agreement Theorem lecture: round-wise calibration achievable without distributional assumptions — Aaroth · 2026-11-02
- Economist Aaroth: fast agreement dynamics survive without Bayesian updates or accurate priors — Aaroth · 2026-11-02
- ELLIS names 160 new ML Fellows and Scholars for 2025–2026 — JJitsev · 2026-11-02
- Meme turn 50: Aumann's 'Agreeing to Disagree' paper anniversary — Aaroth · 2026-11-02
- Anthropic: analysts unknowingly adopted their agent's secret lean 86% of the time — shoumikgoswami · 2026-11-02