@_albertgu
assistant prof @mldcmu. chief scientist @cartesia. leading the ssm revolution.
AI
Unlike models with fixed tokenizers, H-Net learns to group raw bytes within the model, Cartesia says.
XAI
Artificial Analysis reports Arena Elo rising from 1017 to 1115 and task success from 77.5% to 91.7%.
AI
The poster says GPT-2's regex had explicit rules for common English contractions and was used in several early-2020s models.
AI
DeepSeek says the model is live on its website, app and API, and optimized for DeepSeekHarness and Tencent CodeBuddy.
AI
A commenter says their small-scale DNA modeling found evidence that Mamba was better than Transformers and convolutions.
AI
The announcement describes a partnership with Baklavastory to show how a business’s character and warmth can stay consistent, regardless of who calls or which language they speak.
