Search papers, labs, and topics across Lattice.
5
0
10
Current AI agents only manage to complete 20.6% of complex real-world tasks, revealing a stark gap in their capabilities compared to human users.
Achieving an ultra-low specific surface area of 1.89 m2 g-1, this hard carbon anode delivers a remarkable reversible capacity of 342.3 mAh g-1 with an initial Coulombic efficiency of 90.4%.
Limited human oversight can dramatically enhance the safety and stability of self-evolving agents, challenging the notion that autonomy must come without guidance.
Forget hand-crafted benchmarks: CUA-Gym's auto-generated training data lets computer-use agents crush existing open-source models on real-world tasks.
Forget scaling *out* -- TeraPool scales *up* to 1024 RISC-V cores sharing L1 memory, achieving impressive TFLOP/s and energy efficiency thanks to its hierarchical interconnect.
