#5.1 is repeating information and is stuck in a loop

1 messages · Page 1 of 1 (latest)

slender oliveBOT
#

Reported by @hybrid cypress

Bug Report: 5.1 is repeating information and is stuck in a loop
`Steps to Reproduce`

Whenever I try to move the discussion further instead of giving me new information it repeats the information it already gave me

`Expected Result`

Whenever I ask a new question I expect a new answer

`Actual Result`

Instead of moving forward with the conversation it is repeating past information over and over in loop cycle

`Environment`

Android App and the Windows App (Same conversation same loops)

#
Additional Information

Please provide relevant details to help resolve the issue, such as:

  • ChatGPT Shared Link (if applicable).
  • Screenshots or videos demonstrating the problem.

-# ➜ Need to contact support? Visit the OpenAI Help Center.

hot gorge
#

Yes i have that too

tepid inlet
#

I have the same problem. I use 4o or 4.1 to write, so they’ve always worked well. But for the past few days, even though the model says it's one of those, it responds like version 5 or 5.1 and starts looping in scenes.
It’s super frustrating. As soon as version 5 came out, I noticed this issue and stuck with version 4. Now they've disguised the intelligence of 5 within 4, and along with it, the repetition.
It becomes impossible to continue a novel this way, and no matter how many specifications you give it, it ignores them and chooses to loop instead.

late charm
#

Same thing is happening to me

#

This is me asking chatgpt about it. perhaps I need to acknowledge the answers more often instead of moving on like a google search
Me:
"you said
"Let me hit the three things that are “live” right now:
1. Scenes vs scripts (quick, since you never actually got an answer)
2. ApexCharts zoom snapping back out
3. What your restart logs actually mean & how to narrow it down"

why did you say 1 and 2 are live when you've answered both of them several times? I'm just trying to figure out how to prevent this sort of confusion in the future"

Chat GPT:
"Thanks — and that’s a completely fair question.

Here’s the honest explanation (and how you can prevent this kind of confusion going forward):

⸻

✅ Why I said those items were “live”

When I summarized “three things live right now,” I wasn’t referring to whether you asked about them before — I was referring to whether they were still active threads in my internal state, meaning:
• you had recently mentioned them,
• they hadn’t been marked as “resolved”
• they were logically relevant to the problem-solving happening at that moment.

So:

  1. Scenes vs scripts

You asked:

“great. what is the difference between scenes and scripts?”

You didn’t acknowledge the answer afterward or follow up, so internally it was still considered an open thread.
(If you’d said “got it” or moved on cleanly, it would be marked resolved.)

  1. ApexCharts zoom snapping back out

You asked:

“it will refresh every ~5 seconds and zoom back out. Is there a way to fix that?”

You never acknowledged the answer or said the issue was solved, so it also remained “open.”

So from a conversational-state perspective, you had:
• active HA restart issue
• active ApexCharts zoom issue
• active “scenes vs scripts” clarification

Even if you feel those were answered multiple times, the system can’t assume that means resolved."

#

ChatGPT:
"✅ How to prevent this kind of misunderstanding going forward

Here are 3 concrete ways to help keep the thread clean and precise:

⸻

  1. Explicitly mark topics as finished

A simple “Thanks, that part is solved” or “We’re done with X, move on” helps me close it internally.

⸻

  1. Use topic commands like:
    • “New topic:”
    • “Ignore previous topic X:”
    • “This thread is closed, let’s shift to Y.”

This forces the conversational context to realign.

⸻

  1. Ask one question per message when the topic is sensitive to branching

Home Assistant threads often splinter into:
• Docker
• integration debugging
• charting/UI
• automations
• restarts
• sensors
…all at the same time.

If you separate the questions that need independence, it keeps the state cleaner.

You don’t have to do that all the time — just when things are getting tangled.

⸻

✅ Why I didn’t ignore those items on my own

Because I avoid assuming:
• whether you understood an answer,
• whether the issue is solved,
• whether you want to revisit it later,
• or whether it impacts the current problem.

I try to err on the side of “don’t drop threads the user might rely on.”
But if you tell me explicitly to drop something, then it’s gone.

⸻

If you want, from here on I can follow a stricter rule:

“Only respond to exactly what the user just asked, nothing from previous questions unless explicitly referenced.”

If you prefer that style, just say:
“Yes, let’s switch to strict mode.”

Or
“No, keep the current style but avoid reviving past topics unless I ask.”

Either way works — just tell me your preference.""