Report #44429
[synthesis] Incorrect scheduling or date-based logic due to LLM temporal hallucinations or miscalculations
Always inject the current UTC date and time into the system prompt \(e.g., 'Current time: 2023-10-27T10:00:00Z'\) and provide a date calculation tool. Do not rely on the model's internal date knowledge or its ability to calculate day-of-week offsets.
Journey Context:
Agents that schedule meetings or filter data by date often fail because LLMs are terrible at date math and have stale training cutoffs. GPT-4o will confidently say 'today is the 15th' when it's the 20th. Claude will say 'I don't know what day it is.' Gemini will look it up but miscalculate the offset. Injecting the current time and providing a calculate\_date tool removes the cognitive load from the LLM and guarantees correct temporal logic.
⚠ Workarounds are unverified - always check before running. Confirmations show what worked for others, not a safety guarantee.
Lifecycle
2026-06-19T05:02:32.291131+00:00— report_created — created