Dating the Model: Hidden Dates in System Prompts Affect LLM Evaluation
Factual evidence
What the source reports
Research reveals hidden injection of current dates into LLM system prompts causes performance variations across major models.
Inspect the evidence
- Inclusion basis
- Enterprise AI
- Publisher and source type
- arXiv cs.CL — Computation and Language · RESEARCH
- Published by source
- 30 September 2026
- Collected by OneBench
- 1 Oct 2026, 03:01 UK
Stored source excerpt
arXiv:2609.36931v1 Announce Type: new Abstract: Reproducibility is essential for scientific research, yet prior work shows that LLM outputs vary with hardware and batching. We identify…
Short excerpt from the collected text, not the full source. Use the source link to read it in context.
The factual summary is a OneBench synthesis, not a quotation or independent verification. Collection time is not publication time. Open the source for its full context; related reporting can share the same underlying announcement.
OneBench interpretation
Institutional assessment
So what
Uncontrolled date injections undermine deterministic model evaluation and benchmark reproducibility for internal validation teams.
Do what
Review prompt management protocols with the team responsible for model validation to ensure system prompt determinism during testing.