Search Results for "decomposition"

...The core idea is to enable agents to generalize learned behavior by representing continuous control policies in a compact, interpretable form that preserves smoothness and controllability. The project includes implementations that experiment with policy encoding, action decomposition techniques, and sample efficiency analysis in classic reinforcement learning environments. By organizing action outputs into meaningful, lower-dimensional manifolds, MapCN attempts to improve both learning convergence and transfer performance when compared to unguided continuous control baselines.

Dexter is an autonomous agent tailored for deep financial research: you pose complex financial questions (for example, about a company’s revenue growth or financial ratios) and Dexter breaks them down into structured research tasks, fetches relevant real-time data (e.g. income statements, cash flows), performs analysis, and returns data-backed answers. It uses a multi-agent architecture with components such as a planning agent (to decompose queries), an action agent (to run tasks & fetch...