Search papers, labs, and topics across Lattice.
7
0
8
0
OPD-Aha is introduced, which reconstructs the distillation target directly from this isolated visual preference rather than relying on the fragile teacher-student discrepancy, yielding broad and consistent improvements across diverse fine-grained perception and complex multimodal reasoning benchmarks.
ECLIPSE achieves up to 96.7% attack success against long-horizon agentic systems, revealing critical vulnerabilities in current defenses.
A single email can stealthily poison the memory of personal agents, leading to long-term manipulation of their behavior without detection.
Self-evolving LLMs can amplify adversarial threats, making every known attack lineage-persistent and exposing critical vulnerabilities that static defenses can't address.
Trustworthy memory search can significantly reduce vulnerabilities in personal AI agents without compromising their long-term memory utility.
MCP-based AI systems are alarmingly vulnerable to caller identity confusion, allowing unauthorized access to sensitive tools and operations after just one initial authorization.
Embodied AI's failures aren't just LLM or CPS problems; they arise from embodiment itself, where safe-seeming decisions cascade into real-world chaos.
