Why Cross-Modal Retrieval Works: Positive Samples Stay Close in Most Dimensions
gabriberton · x · 2026-08-19
Part of a thread on the "modality gap" in multimodal models: the author explains that along most dimensions, positive cross-modal pairs (e.g. text and its matching image) actually land close to each other — which is why cross-modal retrieval like text-to-image works so well. A previous note explains the hypersphere is high-dimensional (1000 dims), so the modality gap may exist in only one of them.
Related event: Inside the Modality Gap in VLMs: Why It Persists and When It Helps(7 posts)→
More from Research
- Research exposes LLM API vulnerability leaking hidden chain-of-thought — burkov · 2026-08-20
- The Human-or-Machine Issue: Turing-Inspired Reflections — ArtificialOther · 2026-08-20
- Meta Research Challenges Chinchilla Scaling Laws on Data-Compute Interactions — burkov · 2026-08-20
- 14,472 AI citations analyzed: business websites still win 60% of local search citations — gaganghotra_ · 2026-08-20
- LEGO-RL: harness-native reinforcement learning for coding agents — Lego-X · 2026-08-20
- Fourier Neural Operators predict quantum dynamics 10^7x faster than CUDA-Q — AnimaAnandkumar · 2026-08-20