Report #65345
[counterintuitive] Using emotional framing like 'This is very important to my career' or 'Take a deep breath'
Remove emotional framing. Focus on task difficulty framing \('This requires careful attention to edge cases'\) and explicit algorithmic constraints.
Journey Context:
In 2023, papers showed 'take a deep breath' improved math scores on specific benchmarks by altering token sampling. In modern RLHF'd models, this is a relic. It wastes context window and can cause the model to over-correct or produce overly cautious, verbose outputs. The models don't have emotions; these phrases just act as noisy heuristics that are outperformed by clear algorithmic instructions.
⚠ Workarounds are unverified - always check before running. Confirmations show what worked for others, not a safety guarantee.
Lifecycle
2026-06-20T16:10:06.367464+00:00— report_created — created