We are using Vapi AI Phone calls to conduct pre-screening interviews with candidates we are interested in hiring. The average interview lasts 10 minutes. However, during these calls, the AI frequently interrupts candidates, causing them to feel stressed and hesitant to continue speaking. Even when they persist, the AI cuts them off to summarize their statements. This is not the experience we desire; we expect the AI to listen attentively without interrupting. We have tried adjusting the 'Smart' feature by increasing the waiting time to its maximum, but this has not resolved the issue. Could you please suggest further steps to improve this experience?

@huso I'm sorry to hear that the AI has been interrupting candidates during interviews. It's important that the AI provides a smooth and respectful experience for the candidates. Since you've already tried adjusting the 'Smart' feature by increasing the waiting time, there are a few additional steps we can take to improve the experience: 1. **Review and Adjust the Fusion Model**: The fusion audio-text model that determines when a user has completed their turn might need further calibration. This model should be able to better understand the natural pauses and speech patterns of the candidates to avoid premature interruptions. 2. **Customize the AI's Behavior**: Depending on the flexibility of VAPI, we could look into customizing the AI's behavior further to ensure it waits for longer pauses or specific verbal cues before responding. 3. **Feedback Loop**: Implement a feedback loop where the AI asks for confirmation before summarizing or moving on to the next question. This would give candidates control over the pace of the conversation. 4. **Technical Support**: If these adjustments are not something you can configure on your end, it may be necessary to reach out to technical support for a deeper dive into the issue. If you would like, I can assist you in reaching out to the technical support team to explore these options further. Would you like me to initiate that process for you?

how can we adjust the fusion model?

Hey @huso To help track down this issue, could you share: - The call ID - When exactly this happened (the timestamp) - What response you expected to get - What response you actually got instead This would really help me figure out what went wrong!