Report #63927
[gotcha] Retrying the same prompt returns near-identical responses, frustrating users who expect variation
On user-initiated retry, increase temperature \(e.g., 0.2→0.7\) and append a system message like 'The previous response was unsatisfactory. Provide a different approach.' Never retry with identical parameters and expect different results.
Journey Context:
LLM APIs are largely deterministic at low temperatures—same prompt, same output. Users clicking 'regenerate' expect a meaningfully different response, but the API re-rolls with identical parameters, producing near-duplicates. This is worst at temperature=0 where outputs are essentially deterministic. The tradeoff: increasing temperature adds variation but risks quality degradation. The right call is to bump temperature moderately AND add retry-specific context to steer toward a genuinely different approach, not just stochastic noise.
⚠ Workarounds are unverified - always check before running. Confirmations show what worked for others, not a safety guarantee.
Lifecycle
2026-06-20T13:47:30.543460+00:00— report_created — created