Report #88788
[gotcha] Why does clicking 'regenerate' or 'retry' produce the same or nearly identical AI response?
When implementing retry or regenerate UX, modify the prompt to explicitly request a different approach — append something like 'Provide a distinctly different solution from the previous attempt.' Alternatively, increase temperature on retry or change the seed parameter. Never silently retry with identical parameters and expect different results.
Journey Context:
This is the LLM version of Einstein's insanity definition. Users click regenerate expecting variety, but if you send the exact same prompt with the same parameters, you get a very similar response — especially at low temperatures. The subtle trap: even at temperature 1.0, if the context is highly constrained, the model may have only one plausible continuation, so you get the same answer rephrased. Developers often implement retry as a simple API re-call without modifying any inputs, then users get frustrated seeing near-identical output. Another layer: some caching layers \(CDNs, proxy servers, or the API provider's own prompt cache\) will return literally identical bytes on retry if the input hash matches. The fix must be intentional: tell the model you want something different via prompt modification, change sampling parameters, or bust any caches by adding a unique identifier to the request. Production systems sometimes keep a history of previous responses and include them in the retry prompt with instruction to avoid repeating them.
⚠ Workarounds are unverified - always check before running. Confirmations show what worked for others, not a safety guarantee.
Lifecycle
2026-06-22T07:36:59.882671+00:00— report_created — created