Within-agent score std exceeds cross-seat win-rate differentials by 1–2 orders of magnitude

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

• Within-agent score standard deviation suggests deck order matters more than seat position.claim0.820
  Observation from variance analysis, though not tested as hypothesis.
• G3-F within-agent score std 4,026 on median 5,250finding0.757
  High score variance driven by deck order.
• Higher-density priors (B10) are more robust to cross-base OOD drops than lower-density ones (B9) after fine-tuningfinding0.725
  E2 asymmetric transfer finding consistent with UCCT's mismatch-driven OOD fragility
• Coherence scores remain between 60-70 across all cross-domain steering combinations in OLMo-3finding0.721
  Shows persona steering largely preserves response quality across evaluation domains
• Within-cell standard deviation of trait-expression scores: mean 15.11, median 15.76, min 0.00, max 44.22 on 0-100 scale; grows with coefficientfinding0.715
  Documents heterogeneity of steered responses; steering increases response variance
• Nearly all of 64 simulated agents attain 100% performance at around the 10th trial when allowed to perform abductive Bayesian model reduction after each trial.finding0.715
  Group-level simulation result showing generalizability of BMR benefit across agents
• Under AS, Llama-3.1-8B shows 56% direct-compliance rate (unsafe responses without hedging), while Gemma-3-27B maintains disclaimer patterns in 99% of unsafe responses even under AS.finding0.714
  Qualitative failure mode difference between architectures under activation steering
• Ablating 26 OTD latents reduces multi-attempt rate by 25% (from 7.4% to 5.5%) in Llama-3.3-70Bfinding0.712
  Primary causal evidence for dedicated internal consistency-checking circuits