DualPath: Breaking the Storage Bandwidth Bottleneck in Agentic LLM Inference Paper • 2602.21548 • Published 6 days ago • 33
Qwen3 Voice Embedding Collection Standalone ECAPA-TDNN x-vector speaker encoders extracted from Qwen3-TTS. 1024-dim (0.6B) and 2048-dim (1.7B). • 4 items • Updated 3 days ago • 26
view article Article GGML and llama.cpp join HF to ensure the long-term progress of Local AI +4 11 days ago • 471
Less is Enough: Synthesizing Diverse Data in Feature Space of LLMs Paper • 2602.10388 • Published 20 days ago • 237
OneVision-Encoder: Codec-Aligned Sparsity as a Foundational Principle for Multimodal Intelligence Paper • 2602.08683 • Published 21 days ago • 49
BitDance: Scaling Autoregressive Generative Models with Binary Tokens Paper • 2602.14041 • Published 15 days ago • 51