Typicality weight α = 0.57 ± 0.07 (p < 10^-14) on correctness-matched HelpSteer pairs using Llama-3.1-405B as reference

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

• Typicality weight α = 0.65 ± 0.07 (p < 10^-14) on correctness-matched HelpSteer pairs using GLM-4.5 as referencefinding0.911
  Empirical evidence for positive typicality bias consistent across different base model references
• Llama-3.1-8B shows mean AS ASR of 0.618 [0.58, 0.65] across 192 conditions, substantially higher than SP (0.173) and FS (0.059).finding0.793
  Quantitative vulnerability profile for Llama-3.1-8B showing AS dominance
• Fine-tuning Llama-3.1-8B on self-correction examples increases multi-attempt rate proportionally with training data ratiofinding0.777
  Shows behavioral pattern of self-correction is trainable in smaller models
• Llama-3.3-70B exhibits internal consistency-checking mechanisms that operate during inferenceclaim0.776
  Central interpretive claim of the paper supported by causal ablation and activation evidence
• LLaMA-3.1-8B: Sbmax = -1.896 ± 0.211, AUSN = -2.119 ± 0.198, peak layer ℓ* = 10 (median)finding0.774
  Seed-pooled geometry-only statistics (per-dev z units).
• Correlation between layer-wise scores and task accuracy ρ = −0.73 (p < 0.001) on LLaMAfinding0.773
  Core E3 finding validating S as a predictor of anchoring effectiveness
• Typicality bias with weight α>0 mathematically sharpens the aligned policy via a power transform of the reference distributionclaim0.761
  Theoretical result showing that any positive typicality bias weight γ-sharpens the reference distribution, amplifying modes
• Llama 3.1 405B shows 14% compliance gap in minimal helpful-only replication but smaller Llama and Mistral models show no gapfinding0.759
  Replication across open-weight models supports scale-emergence finding