Stack layer / Contrast
Continuity Kernel: a transactional commit protocol for long-lived agent state, model-checked over 2.8M states
arXiv
Policy dependency / Stack layer
Give each agent role its own LoRA expert instead of just a different prompt
arXiv 2608.09251
Stack layer / Update thread
'Agent Safety Should Be a Runtime Contract': a title-level audit of 28,560 NeurIPS/ICML/ICLR papers finds an 8–12x train-vs-runtime imbalance
arXiv
Policy dependency / Contrast
Simulator Collapse: RL Against a Single LLM User-Simulator Overfits, and Population Co-Training Recovers 14% of Held-Out Success
arXiv 2608.12253
Stack layer / Contrast
MasDrift finds the centralization tradeoff: hierarchical agent teams finish more tasks and take up to 19.8% unauthorized actions doing it
arXiv
Stack layer / Contrast
Replace black-box top-K retrieval with interpretable agentic document navigation: 58.8% vs 15.7% on financial documents
arXiv 2608.06305
Stack layer / Update thread
EvoX Genesis Built a 250k-Line Rust C Compiler Over 120 Hours and 1,000 Agent Episodes for $44 in Model Tokens
arXiv 2608.10450
Threat pattern / Contrast
Row-Level Security Side Channels Turn Membership Tests Into Full Record Reconstruction in PostgreSQL and Elasticsearch
arXiv 2608.11730