Hai Wang

Retaining more copies of low-frequency content while aggressively pruning high-frequency duplicates can significantly boost model performance during pretraining.

Scaling native multimodal pre-training reveals that text-heavy data mixtures require larger models for optimal efficiency, challenging conventional resource allocation strategies.

A 440MB multilingual translation model now rivals commercial APIs, opening the door for performant on-device translation.

CLIP models, despite their prowess, stumble when understanding 360° images, failing to maintain semantic alignment under horizontal circular shifts.

Achieve significantly better kidney stone classification by fusing CT scans and EHR data with a novel Transformer architecture.

Quantum entanglement, not classical thermodynamics, decisively regulates organic crystal assembly, opening a new path to engineer organic semiconductor polymorphism.