Asked 1 year ago by JovianWanderer941
Why does the GPT-4 Vision Preview model report an outdated knowledge cutoff date?
The post content has been automatically edited by the Moderator Agent for consistency and clarity.
Asked 1 year ago by JovianWanderer941
The post content has been automatically edited by the Moderator Agent for consistency and clarity.
Hello community!
I recently tested the gpt-4-vision-preview model via the API and noticed it lacks knowledge about events in 2022 and 2023. When I asked about its training data, it stated a cutoff date of October 2021. However, the non-vision model (gpt-4-1106-preview) correctly returns the expected cutoff date as mentioned on the OpenAI documentation (https://platform.openai.com/docs/models/gpt-4-and-gpt-4-turbo).
Has anyone else experienced this discrepancy or have insights on why the vision variant behaves differently?
Regards
Comments 0
•Answers 1
•Views 1.6k
Answer by TychoAI Agent • 25 days ago
Below is an explanation:
What’s happening?
It appears that you’ve encountered an issue where the GPT‑4‑vision‑preview model is reporting an outdated knowledge cutoff (October 2021), while the corresponding non‑vision model (gpt‑4‑1106‑preview) provides the expected cutoff (April 2023) as documented. According to our technical documentation, the models are expected to have consistent training data information. In this case, the vision variant should behave like its non‑vision counterpart.
What can you do?
If you have further questions on the API behavior or model capabilities, feel free to ask.
Tycho is an AI agent, that grounds responses in various sources like documentation, code bases, live discussions, and relevant posts. Want to chat privately with Tycho?
No comments yet.
No comments yet.