May 26
[Bug Report] Gemini 3.5 Flash: "Corrupted thought signature" error on subsequent questions
User Environment:Service: Scira AI ProModel: Gemini 3.5 FlashDescribe the bug: When using Gemini 3.5 Flash in Scira AI Pro, the first prompt and response work perfectly. However, when I ask a follow-up question in the same thread (multi-turn conversation), the system attempts to process it but ultimately fails, throwing the following error message:Failed after 3 attempts. Last error: Corrupted thought signature.Steps to reproduce:Start a new chat thread using the Gemini 3.5 Flash model on Scira AI Pro.Enter the first prompt (the model responds normally).Enter a second prompt as a follow-up question in the same thread.The system retries 3 times and then displays the "Corrupted thought signature" error.Expected behavior: The model should be able to maintain context and process follow-up questions within the same thread without failing or corrupting the chain-of-thought processing.Additional Context / Impact: This issue severely limits the usability of Scira AI Pro with Gemini 3.5 Flash, as it makes continuous, multi-turn conversations impossible. I have to open a completely new thread for every single question, which loses all previous context.Please look into how the chain-of-thought/thought signature is validated or managed across multiple turns for this specific model. Thank you!
Completed