RSS 中文
⌕

INDEX // #KV-CACHE

SYSTEM // ACTIVE // AGGREGATED TELEMETRY FOR ECOSYSTEM NODE

NEWS // Latest Activity TOTAL: 010
Gemma 4's "Divergent" Edge Architecture: A Systems Engineer's Breakdown of Memory Wall Breakthrough for Local AI
Gemma 4's "Divergent" Edge Architecture: A Systems Engineer's Breakdown of Memory Wall Breakthrough for Local AI
#GEMMA#EDGE AI#MEMORY MANAGEMENT
Zhipu AI Reveals "Scaling Pain": Addressing GLM-5 Coding Agent Anomalies Under High Load with Robust System Engineering
Zhipu AI Reveals "Scaling Pain": Addressing GLM-5 Coding Agent Anomalies Under High Load with Robust System Engineering
#KV CACHE#SPECULATIVE DECODING#GLM
Exploring LLM Architectures: KV Sharing and Compressed Attention
Exploring LLM Architectures: KV Sharing and Compressed Attention
#KV-CACHE#ATTENTION-MECHANISM#LLM-ARCHITECTURE
River-LLM Introduces KV-Shared Exit to Achieve Seamless Early Exit and Significant LLM Inference Speedup
River-LLM Introduces KV-Shared Exit to Achieve Seamless Early Exit and Significant LLM Inference Speedup
#LLMS#EARLY EXIT#KV CACHE
DeepSeek's Price Cuts Ignite AI Token Market Reshuffle: Technology Innovation Drives Industry Transformation
DeepSeek's Price Cuts Ignite AI Token Market Reshuffle: Technology Innovation Drives Industry Transformation
#DEEPSEEK#TOKEN PRICING#MOE ARCHITECTURE
SinkRouter: A Novel Training-Free Routing Framework for 2x Faster Long-Context Decoding in LLMs/LMMs
SinkRouter: A Novel Training-Free Routing Framework for 2x Faster Long-Context Decoding in LLMs/LMMs
#LONG-CONTEXT DECODING#ATTENTION MECHANISM#KV-CACHE
Together AI Open-Sources OSCAR: 2-Bit KV Cache Quantization
Together AI Open-Sources OSCAR: 2-Bit KV Cache Quantization
#KV-CACHE#QUANTIZATION#TOGETHER-AI
Baidu Open-Sources Unlimited OCR: Continuous Document Parsing Beats DeepSeek
Baidu Open-Sources Unlimited OCR: Continuous Document Parsing Beats DeepSeek
#OCR#MULTIMODAL#DEEPSEEK
Why Your LLM Doesn't Re-Read the Prompt: The Magic of KV-Cache
Why Your LLM Doesn't Re-Read the Prompt: The Magic of KV-Cache
#KV-CACHE#INFERENCE#LLM
SPIN Framework Unifies Sparse Attention with Hierarchical Memory for Scalable Long-Context LLM Serving
SPIN Framework Unifies Sparse Attention with Hierarchical Memory for Scalable Long-Context LLM Serving
#SPARSE ATTENTION#KV CACHE#LLM SERVING