Study finds video LLMs fail at counting object crossings on screen
A new arXiv paper diagnoses a readout failure in video LLMs: they cannot count how many times an object crosses the screen, a deficit linked to missing object permanence, and proposes a repair approach.
Read the full story at arXiv cs.CV ↗
Timeline · 1 report
- 2026-11-10 05:00 · arXiv cs.CV
Counting Events, Not Seeing Them: Diagnosing and Repairing the Readout Failure in Video LLMs