AI performance costs are falling faster than those of any previous technology
2 sources 4 min
Parallel cut research time and cost in half with GPT‑6 Astra
2 sources 2 min
Making Agents More Consistent: Skills Should Form Habits for Repeat Tasks
Jev-Mem: System-One-Controlled Agentic Memory for Efficient AI Agents
dQwen3.5: Hybrid-Attention Diffusion Language Models
1
Higher-order pruning of experts in mixture-of-experts language models
1
SAS: Simple Attention Sparsification via End-to-End Optimization of Context Ranking
2
SAS: Simple Attention Sparsification via End-to-End Optimization of Context Ranking
1
COBRA-Skills: Contextual Bandit-Guided Evolution for Agent Skill Optimization
CoVeR: Coverage-Based Token Pruning for Multi-View 3D Reasoning in VLMs
A*-Thought-V2: Efficient Latent Reasoning via Geometric Dynamics of LLM
Graph Machine: Towards Better Pretraining via Edges
HyQuant: Hybrid-Precision Quantization for LLM Attention
1
Quantization-Aware Healing: a compressed, 4-bit model that outperforms its full-precision original
Jalapeño’s first results show industry-leading speed and efficiency in AI inference
Making Knowledge Distillation Cheap Enough to Run at Scale
1