Cache-to-Cache: Direct Semantic Communication Between LLMs (2025)
1
On-Demand Attention: Language Models Know When to Recall
DeepSeek-v4.1 Flash: Pushing the Limits of KV Cache Compression
15 min1
DeepSeek-V4.1-Flash: Pushing the Limits of KV Cache Compression
1
Long-Lived Characters, Local Inference: Incremental Memory Maintenance for Game NPCs
JustFit: 200K-Token LLM Serving on a 24 GiB Laptop with Just-in-Time State Management
Fathom: Per-Query Read Depth for Sparse Decoding over Offloaded KV Caches
1
New Deepseek model V4.1-Flash cuts memory needs for AI agents
4 min1
DeepSeek AI Released DeepSeek-V4.1-Flash with 1M Context, FP4 KV Cache, and Cross-Layer Attention Reuse
4 min2
TRACE: Trajectory-robust Admission with Evidence Ordering for Efficient GUI Agents
Grouped Value Attention: Efficient KV Caching via On-Demand Key Reconstruction
LatentPort: Beyond KV Cache - Cross-Model Transfer of Recurrent Memory in Hybrid Language Models: A 4B-to-9B Hybrid-State Handoff Without Target Prefix Replay
Forgetting Without Restarting: Execution-State Unlearning for Stateful LLM Agents
1
BeaconKV: Key-Value Cache Compression Guided by Beacon Queries for Efficient Large Reasoning Model Inference
1
ShallowStream: Index Shallow then Answer Deep for Streaming Video Understanding
HyQuant: Hybrid-Precision Quantization for LLM Attention
1