The 26 differentially-activated OTD latents play a causally important role in enabling ESR 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.

• The 25% reduction in multi-attempt rate from OTD ablation suggests additional mechanisms contribute to ESR beyond the identified latentsclaim0.783
  Acknowledges incompleteness of the causal account, suggesting redundant circuits or nonlinear interactions
• Ali et al. 2025 found contrastive activation addition less effective at larger model scale, consistent with ESR in 70Bfinding0.764
  Prior finding from related work that aligns with ESR being strongest in the largest model tested
• Approximately half of the 26 OTD latents show near-zero or negative effect sizes, activating more during on-topic contentfinding0.762
  Reveals that contrastive search yields a heterogeneous set, not all functioning as true off-topic detectors
• We cannot isolate whether ESR reflects scale, architecture, or training procedures in Llama-3.3-70Bclaim0.761
  Epistemic limitation claim acknowledging confounds in the cross-model comparison
• 26 candidate off-topic detector latents identified in Llama-3.3-70B via contrastive searchfinding0.758
  Core mechanistic finding identifying specific SAE latents associated with ESR
• ESR exhibits non-monotonic relationship with boost level, peaking around -0.3σ below threshold in Llama-3.3-70Bfinding0.748
  Characterizes the narrow operating window in which ESR can manifest
• Meta-prompt ESR enhancement effects scale with model size across Llama and Gemma familiesfinding0.747
  Suggests underlying self-monitoring circuits must be present for meta-prompting to enhance them
• Mean difference patching on Llama-3-8B layer 10 produces intervened EMD exceeding the natural-natural baselinefinding0.745
  Empirical demonstration that MDVP produces divergent representations in a real LLM