⛵ Vela 2.0
Collection
Towards Open Foundation Routing Models • 5 items • Updated • 20
Dhruvajyoti Sarma
dhruva-sarma
AI & ML interests
None yet
Recent Activity
upvoted a collection about 4 hours ago
⛵ Vela 2.0 reacted to Parveshiiii's post with 🤗 about 4 hours ago
Most deepfake audio detectors are quietly cheating.
They don’t really listen to the speech — they just look at how long the embedding vector is. Once they figure that out, accuracy looks great on paper and falls apart in the wild.
AIRealNet-Audio was built to stop that shortcut.
It forces every feature onto the unit hypersphere (twice) so the model can only use direction, not magnitude. Trained on speech from 100+ different TTS and voice-cloning systems, plus real human recordings under heavy compression and noise.
The result is a detector that actually has to learn the artifacts instead of gaming the feature space.
Model: https://huggingface.co/Modotte/AIRealNet-Audio liked a model 4 days ago
fastino/GLiNER2.5-multi-DecideOrganizations
None yet
upvoted a collection about 4 hours ago
reacted to Parveshiiii's post with 🤗 about 4 hours ago
Post
1377
Most deepfake audio detectors are quietly cheating.
They don’t really listen to the speech — they just look at how long the embedding vector is. Once they figure that out, accuracy looks great on paper and falls apart in the wild.
AIRealNet-Audio was built to stop that shortcut.
It forces every feature onto the unit hypersphere (twice) so the model can only use direction, not magnitude. Trained on speech from 100+ different TTS and voice-cloning systems, plus real human recordings under heavy compression and noise.
The result is a detector that actually has to learn the artifacts instead of gaming the feature space.
Model: Modotte/AIRealNet-Audio
They don’t really listen to the speech — they just look at how long the embedding vector is. Once they figure that out, accuracy looks great on paper and falls apart in the wild.
AIRealNet-Audio was built to stop that shortcut.
It forces every feature onto the unit hypersphere (twice) so the model can only use direction, not magnitude. Trained on speech from 100+ different TTS and voice-cloning systems, plus real human recordings under heavy compression and noise.
The result is a detector that actually has to learn the artifacts instead of gaming the feature space.
Model: Modotte/AIRealNet-Audio
liked a
model 4 days ago
liked a
model 10 days ago
upvoted an article 14 days ago
Article
BERT 101 - State Of The Art NLP Model Explained
britneymuller
• • 12 upvoted 4
articles 28 days ago
Article
Beyond LoRA: Can you beat the most popular fine-tuning technique?
- +2
BenjaminB, sayakpaul, hubnemo, kashif
• • 104 Article
Introducing Mellum2: A 12B Mixture-of-Experts Model by JetBrains
JetBrains
• • 40 Article
What We Learned by Reproducing 2,200 papers from ICML
abidlabs
• • 120 Article
Decoding Strategies in Large Language Models
mlabonne
• • 119 upvoted an article about 2 months ago
Article
Explaining the SDXL latent space
TimothyAlexisVass
• • 56 upvoted 2
articles 3 months ago
Article
Introducing RTEB: A New Standard for Retrieval Evaluation
- +4
fzliu, KennethEnevoldsen, Samoed, isaacchung, tomaarsen, fzoll
• • 149 Article
NVIDIA Nemotron 3 Embed Ranks #1 Overall on RTEB, Advancing Agentic Retrieval
nvidia
• • 60 upvoted an article 8 months ago
Article
We Got Claude to Build CUDA Kernels and teach open models!
- +2
burtenshaw, evalstate, merve, pcuenq
• • 158 liked a
model about 1 year ago
upvoted an article about 1 year ago
Article
Optimizing your LLM in production
patrickvonplaten
• • 23 liked a
Space about 1 year ago
liked a
model about 1 year ago
updated a
collection about 1 year ago
upvoted a paper about 1 year ago
upvoted an article about 1 year ago
Article
🪆 Introduction to Matryoshka Embedding Models
- +1
tomaarsen, Xenova, osanseviero
• • 221