Berkeley Artificial Intelligence Research (BAIR)

최신 AI 뉴스, 모델 및 출시 정보 Berkeley Artificial Intelligence Research (BAIR). ['GPT-4o mini', 'GRASP', 'LLM', 'PEVA', 'ResNet', 'Transitive RL (TRL)']

공식 사이트 ↗

연구 🇺🇸

PEVA: Whole-Body Motion Guided Ego-Centric Video Prediction

Researchers from BAIR (Berkeley AI) introduced the PEVA (Predicting Ego-centric Video from human Actions) model, which generates next first-person video frames based on past frames and a sequence of human actions specified via 3D pose changes. The model uses an autoregressive conditional diffusion transformer trained on the Nymeria dataset and can predict atomic actions, simulate counterfactual scenarios, and support long video generation up to 16 seconds. PEVA can also be used for planning by evaluating different action sequences based on similarity to a target image.

Berkeley Artificial Intelligence Research (BAIR)

BAIR (Berkeley AI)27.07 · 17:05

연구 🇺🇸

What Does word2vec Actually Learn? Researchers Finally Have an Exact Training Theory

Researchers from BAIR (Berkeley Artificial Intelligence Research) have presented the first complete quantitative theory of word2vec training. They proved that in practically relevant regimes, training reduces to principal component analysis (PCA) matrix factorization, with each learned feature corresponding to an interpretable concept. The theory allows computing all features in advance based on corpus statistics and algorithm hyperparameters.

Berkeley Artificial Intelligence Research (BAIR)

BAIR (Berkeley AI)27.07 · 17:04

더 보기 ↓