ReTerm explores failure-driven RSI, allowing Code Agents to evolve from their own mistakes.

ReTerm optimizes its decision logic by analyzing failed operations of Code Agent to reduce unnecessary actions in complex terminal tasks. The system is based on research from Fudan University and The Chinese University of Hong Kong, converting error experiences into improved training materials through a failure-driven RSI (Risk-Senefit-Importance) evaluation mechanism, enhancing the Agent's task success rate.