Meta-Prompting for ESR Enhancement

Where ideas in this method connect to the rest of the corpus — the same concept, an analogy, or a restatement elsewhere.

• Endogenous Steering Resistance
  ~Top-Down Attentional Control· biology~Attention Schema Theory· cognitive

Entities in the same semantic neighborhood but without a typed relation to this one — candidates for new edges or unrecognized duplicates.

• Meta-prompt ESR enhancement effects scale with model size across Llama and Gemma familiesfinding0.781
  Suggests underlying self-monitoring circuits must be present for meta-prompting to enhance them
• Meta-Reflection Prompts as Drift Triggersconcept0.769
  User messages that push the model to reflect on its own processes reliably cause persona drift away from the Assistant
• The meta-prompting scaling pattern suggests underlying self-monitoring circuits must already be present for prompting to enhance themclaim0.754
  Mechanistic interpretation of why meta-prompting effects scale with model size
• Explicit ESRconcept0.734
  The form of ESR focused on in this paper, measured by verbal self-interruption phrases as segment boundaries
• ESR differs from the Hydra Effect in that ESR involves active, online detection and correction with explicit self-interruption tokensclaim0.727
  Distinguishes ESR from prior work on model self-repair
• How does ESR respond to safety-relevant steering interventions, e.g. toward harmful content?question0.726
  Key open question for AI safety implications of ESR
• Can ESR be adversarially circumvented?question0.725
  Open security question about robustness of ESR-based defenses
• Meta-prompting increases Llama-3.3-70B multi-attempt rate 4.3× (from 7.4% to 31.7%)finding0.716
  Demonstrates ESR can be deliberately enhanced through prompting in the largest model