The check failed. Why did the AI remember the answer?

说明与用途

当检查失败时，旧逻辑会将未验证的答案重新发送到下一轮，导致 AI 记住错误信息。新逻辑阻止此重放，将答案排除在后续上下文之外。

问题

When a check on an AI assistant’s answer fails, the answer may still be stored and sent to the next model call, causing the assistant to “remember” an unverified response.

结果

The patch ensures that answers failing validation are not automatically included in future model calls, preventing the assistant from using unverified information as context.

工作流

• 1. Configure a check that can pass, reject, or throw.
• 2. Run the assistant with a test model that returns a distinctive marker text.
• 3. Observe the conversation history sent to the model after the check outcome.
• 4. Compare old logic (replays answer on error) vs. fixed logic (does not replay).

证据

In our controlled comparison, the old logic raised an error and saved the answer anyway. The next model call received it.

来源

The fixed logic preserves that distinction. It does not declare the answer disproved, and it does not automatically admit the answer as context for later work.

来源