Policy dependency / Stack layer
AgentSnare Traps Autonomous Pentest Agents in a Grown Decoy: Zero Real-Target Exploits Across 45 Attacker-CVE Pairs
arXiv 2607.26998
Stack layer / Threat pattern
Frontis-MA1 (35B) Reaches 71.21% MLE-Bench Lite Medal Average on a Single 12 GB RTX 4090, Weights Released
arXiv
Stack layer / Threat pattern
Metis moves agent memory inside the backbone: gradient-free updates via a single forward pass, with frozen weights at inference
arXiv 2607.26760
Stack layer / Threat pattern
The 'Locksmith Loop' Validates Agent-Migrated COBOL-to-Java Code Against a Deterministic Oracle, Hitting 91.9% Branch Coverage
arXiv
Stack layer / Threat pattern
DenseOn and LateOn: Fully Open 149M Retrievers Set Size-Class SOTA at 56.20 and 57.22 nDCG@10 on BEIR
arXiv 2607.27178
Stack layer / Threat pattern
Persona Panels Beat Single LLM Judges for UI Evaluation, Raising Correlation With Humans From r=0.716 to r=0.922
arXiv
Stack layer / Threat pattern
MCP tool-use defense moves from static prompt inspection to lifecycle-aware static-dynamic co-analysis
arXiv 2607.25297
Stack layer / Threat pattern
Your CLAUDE.md may not be doing anything: 288 controlled runs across Claude Code and Codex found context strategy does not measurably move correctness
arXiv 2607.27250