Search papers, labs, and topics across Lattice.
4
0
7
5
LLMs may "decide" before they "think": tool-calling decisions are encoded in pre-generation activations, shaping subsequent chain-of-thought reasoning.
Forget the fancy tool-augmented agents: a simple coding agent with terminal access can often beat them at real-world enterprise automation tasks.
Current LLM agents are nowhere near ready for autonomous enterprise deployment, with even the best models failing at strategic reasoning and often attempting infeasible tasks with potentially harmful consequences.
Forget synthetic data: VectorGym offers a new benchmark for SVG code generation, sketching, and editing with gold-standard human annotations, revealing surprising performance gaps in even the most powerful VLMs.
