Financial Context-Rich Strategy for Actor-Critic based Agents: A Deep Reinforcement Learning Approach for Stock Trading
Resumo indisponível. Consulte o paper original.
PAPER / ARXIV:2609.09134
Zhou Yu, Bin Bi, Shiva Kumar Pentyala, Shubham Mehrotra, Sougata Chaudhuri, Shilpa Bhagavath, Zeyuan Chen, Ran Xu, Phil Mui, James Zhu, Sitaram Asur
RESUMO
Agent harnesses (system prompt, tool set, execution hooks, context-management scaffolding) are a critical determinant of agentic task success. Across seven enterprise agent tasks, training a weaker model on an expert's complete trajectories under an evolved harness backfires, regressing performance by 4 to 30 points, because imitation disrupts model-harness fit. We develop an on-policy expert-correction pipeline that localizes the failing turn and asks the expert to rewrite only that turn, preserving the model's planning style.
NO MESMO MAPA
Resumo indisponível. Consulte o paper original.
Resumo indisponível. Consulte o paper original.
Resumo indisponível. Consulte o paper original.
Resumo indisponível. Consulte o paper original.