ReAct

“Existing architectures do not instantiate this full stack. For example, standard RAG systems retrieve based solely on keywords without reasoning, while existing search agents such as ReAct interleave thought and action but lack an explicit mechanism to target generated queries toward answer reasoning.”

“There is a fundamental mismatch between the agent's actual execution history and the reshaped prompt presented to the model. This structural blindness masks crucial state parameters; specifically in RAG tasks, it leads to repetitive queries and useless interactions with search engines.”

“ReAct RAG occasionally provides marginal gains over RAG by enabling iterative retrieval and broader evidence coverage; however, it does not consistently translate additional retrieval steps into reliable performance improvements”

“However, a fundamental limitation of ReAct is that the reasoning and retrieval plan exists entirely within the LLM's context window, leading to context overflow as reasoning chains grow, plan fragmentation, and high latency from sequential execution.”

“However, it cannot foresee the features of different retrieval sources and heavily relies on their descriptions for selection, leading to low-quality and unstable multi-source retrieval.”

“However, planning for complex questions is non-trivial, especially for smaller LLMs (with fewer than 10 billion parameters), which often require supervised fine-tuning”