SteerablePlex: האם ניתן לשלוט בדגמים דו-כיווניים?

תקציר מקורי באנגליתarXiv:2610.12201v1 Announce Type: cross Abstract: Full-duplex speech models can listen and speak simultaneously, enabling natural interaction, but become increasingly difficult to control as the conversation history grows. When used as user simulators, this lack of control can cause them to deviate from prescribed scenarios and produce unreliable evaluation outcomes. We introduce SimIF-Bench (Simulator Instruction-Following Benchmark), which evaluates whether a conversational model stays within a prescribed scenario and completes multiple goals in the required order. The benchmark reveals that current open-source full-duplex models struggle to follow such constraints. We then introduce a Group Reward-Decoupled Normalization Policy Optimization (GDPO)-based training recipe that enables a fu