Top Videos
Top Searches
Moods & Genres
Autumn
Chill
Commute
Energize
Feel-good
Focus
Gaming
Hanukkah
Party
Romance
Sad
Sleep
Summer
Winter
Workout
African
Arabic
Blues
Bollywood & Indian
Classical
Country and Americana
Dance and electronic
Decades
Family
Folk and acoustic
German hip-hop
German pop
Hip-hop
Indie and alternative
J-Pop
Jazz
K-Pop
Latin
Mandopop & Cantopop
Metal
Pop
R&B and soul
Reggae and caribbean
Rock
Schlager
Soundtracks and musicals
Search music and videos
[alphaXiv] Coding Agents Now Build GPU Kernels 4.4× Faster From Scratch
05:30
|
54 views
|
Download MP3
Related Videos
[Hugging Face] Reinforcement Learning Can Merge Experts Without New Rollouts
[alphaXiv] AI Agents Can Now Run Complex Scientific Simulations Reliably
[alphaXiv] Video Demonstrations Can Teach Robots to Act With Less Data
TabPFN-3.5 Just Beat 89 Tabular Models—Even on Messy Data
[Papers with Code] A New Robot Touch Model Boosts Dexterous Success by 25 Points
[Hugging Face] Qwen3-VL-8B-Instruct Can Watch Live Video and Know When to Act
RenderFormer-V2: Neural Rendering at 100K+ Primitives—No Per-Scene Training
[alphaXiv] Large Language Models Can Use Short Contexts—and Decode 7x Faster
[Papers with Code] AI Builds Tiny, Interpretable Transformers Without Training
[Papers with Code] Full-Duplex AI Can Listen, Watch, and Interrupt You in Real Time
[Papers with Code] DiscoveryRSI: AI Agents That Learn to Discover Better Hypotheses
Efficient, Universal, and Better-Conditioned Neural Architectures
[Hugging Face] Text-to-Speech Just Ditched Audio Tokenizers—Without Losing Quality
Why Neural Networks Learn in Sudden Jumps—and Follow Scaling Laws
[alphaXiv] Reinforcement Learning Builds Controllable World Models with 85× Less Data
Model Extraction From Hard Labels Can Rebuild Deep Networks
This Tiny Neural Decoder Reaches 63 dB With Just 594K Parameters
In-Context Sampling: Transformers Generate New Data Without Training
Spectral Optimizers and Stability: Robust Learning Theory for Modern Training