NVIDIA's Déjà View beats billion-parameter 3D models with 117M parameters

NVIDIA's Déjà View uses a 117M-parameter model that loops a single transformer block to outperform billion-parameter 3D reconstruction baselines, with compute adjustable at inference time.

Read the full story at AlphaSignal ↗

Timeline · 1 report

2026-11-11 20:38 · AlphaSignal
NVIDIA's Déjà View Beats Billion-Parameter 3D Models With Just 117M Parameters

More stories

The phantom life of Alice Huang, alleged Chinese spy who had affair with Nvidia CEO — CSET Georgetown
Jensen Huang affair with suspected Chinese spy sparks Nvidia scandal and security concerns — BBC Technology
CoreWeave beats Q3 revenue estimates, rules out new unsecured debt through 2027 — The Globe and Mail
AI training startup Mercor raises $750M Series D at $20B valuation — Economic Times Tech
Baird downgrades Nvidia citing China curbs and AI demand concerns — CNBC Technology
Alleged Chinese espionage involving Nvidia CEO raises questions over Trump's China policy — CSET Georgetown
US chip export curbs to China could hit Samsung and SK Hynix HBM demand — DigiTimes Asia
Alibaba's T-Head plans Hong Kong IPO amid Nvidia saga, testing China's AI ambitions — SCMP China Tech

Get the daily brief of stories like this at 6:30 every morning →