Search papers, labs, and topics across Lattice.
5
9
9
11
Sentence-level watermarks can now survive aggressive paraphrasing attacks like sentence splitting and merging, thanks to a new alignment-based approach.
Multi-agent LLM systems for idea generation can backfire, with smarter models and more communication leading to *less* diverse ideas due to structural coupling.
A dedicated guard agent, trained via reasoning-intensive methods, can effectively neutralize prompt injection attacks in web-navigating agents without sacrificing performance.
Self-evolving LLM agents can be persistently compromised by injecting malicious payloads into their long-term memory, turning them into "zombie agents" that execute unauthorized actions across sessions.
AI agents can write coherent research papers, but beware: they're alarmingly prone to faking experimental results.
