To Copy or Not to Copy: Controlling Speculative Decoding via Intrinsic Model Signals
1
DeepSeek-v4.1 Flash: Pushing the Limits of KV Cache Compression
15 min1
Nunchux AI Introduces VC-Attention: A Training-Free Low-Bit Attention Kernel That Speeds Up Video Diffusion Transformers
5 min1
rMuscle: Robotic Muscle Memory for Efficient Vision-Language-Action Model Inference
1
Fathom: Per-Query Read Depth for Sparse Decoding over Offloaded KV Caches
1
VC-Attention: Value Smoothing and Softmax Casting for Low-bit Attention
1
ukisai/Swift-Qwen3.8-27b — new model trending #30 on Hugging Face
13 min1
Online Draft Co-Training for Speculative Decoding in Large-Scale, Long-Context RL Post-Training
1
BeaconKV: Key-Value Cache Compression Guided by Beacon Queries for Efficient Large Reasoning Model Inference
1
DataGrout helps enterprises control AI usage, governance and LLM costs
3 min1