ESR exhibits non-monotonic relationship with boost level, peaking around -0.3σ below threshold in Llama-3.3-70B

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

• All five judge models consistently rank Llama-3.3-70B as having substantially higher ESR rates than other modelsfinding0.817
  Cross-judge validation of the primary ESR finding across OpenAI, Alibaba, Anthropic, and Google judge models
• LLaMA E3 geometry summary: S_max = −1.896 ± 0.211, AUS_N = −2.119 ± 0.198, peak layer ℓ* = 10 [IQR 0.384]finding0.780
  Seed-pooled geometry statistics for LLaMA in E3, providing quantitative basis for geometry-to-behavior correlate
• Meta-prompt ESR enhancement effects scale with model size across Llama and Gemma familiesfinding0.775
  Suggests underlying self-monitoring circuits must be present for meta-prompting to enhance them
• We cannot isolate whether ESR reflects scale, architecture, or training procedures in Llama-3.3-70Bclaim0.775
  Epistemic limitation claim acknowledging confounds in the cross-model comparison
• Ali et al. 2025 found contrastive activation addition less effective at larger model scale, consistent with ESR in 70Bfinding0.773
  Prior finding from related work that aligns with ESR being strongest in the largest model tested
• Multi-attempt improvement rate peaks at 83% around -1.0σ below threshold in Llama-3.3-70Bfinding0.768
  Shows slightly weaker steering allows more successful corrections, characterizing optimal ESR conditions
• LLaMA-3.1-8B: Sbmax = -1.896 ± 0.211, AUSN = -2.119 ± 0.198, peak layer ℓ* = 10 (median)finding0.767
  Seed-pooled geometry-only statistics (per-dev z units).
• Random latent ablation produces slight increase in ESR rate (3.8% to 4.2%), not statistically significantfinding0.755
  Control result confirming OTD ablation effect is specific to those latents, not a general ablation artifact