Cracking Language Model Memory: The Rise of FARMA and SENTINEL
Persistent memory in large language model agents opens the door to new vulnerabilities. FARMA exploits this by poisoning reasoning histories, but SENTINEL offers a solid defense.
Read the full articleYou might also wanna read
Memori launches agent-native persistent memory infrastructure using structured knowledge graphs from agent trace data
Memori launched its new agent-native memory infrastructure, enabling agents to create structured, long-term memory directly from agent trace
Product Hunt·2mo ago
LLM Memory Architecture: Trade-offs and Implementation Strategies for Production AI Agents
Learn how to architect persistent, scalable memory into AI systems with this technical breakdown of LLM memory types and failure modes.
blog.n8n.io·15d ago
OWASP Agent Memory Guard: Open-source runtime defense against AI agent memory poisoning attacks
AI agent memory security gets a runtime layer: OWASP's Agent Memory Guard screens every write, hitting 92.5% recall at 59µs latency.
helpnetsecurity.com·1mo ago
Deployment-Time Memorization in Foundation-Model Agents: Privacy-Utility Tradeoffs in Persistent Memory Systems
Foundation-model agents are increasingly long-lived systems that remember users across interactions, making memorization an explicit deploym
arxiv.org·26d ago
A Software-Cache Architecture for Human-Supervisable Agent Memory: Coherence Protocol for Persistent Agent Fleets
A software-cache architecture for human-supervisable agent memory. - treaves-GSD/soft-cache-agent-indexing
github.com·16h agoA Software-Cache Architecture for Human-Supervisable Agent Memory: Coherence Protocol for Persistent Agent Fleets
A software-cache architecture for human-supervisable agent memory. - treaves-GSD/soft-cache-agent-indexing
github.com·16h ago
JAMEL: A Framework for Joint Memory and Exploration Learning in Language Model Agents
In open-ended environments, exploration is fundamental for autonomous agents, yet current language model agents struggle with this. Effectiv
arxiv.org·1mo ago
Comments
Sign in to join the conversation.
No comments yet. Be the first.