Two-shot redefinition of "−" operator flips model output from -1 to 23 on 15-8=?

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

• 2-shot reinterpretation of '-' yields 23 for 15-8 on held-out queryfinding0.771
  E1 qualitative: two exemplars (2-3=5, 7-4=11) cause LLMs to output 23 for 15-8.
• Two exemplars (2−3=5, 7−4=11) induce reinterpretation of '−' as addition on held-out queries across mainstream LLMsfinding0.745
  E1 qualitative finding demonstrating anchor rebinding of strong arithmetic prior
• S = ρd - dr - log k is a predictive correlate of when few-shot behavior flipsclaim0.733
  Claim that S predicts threshold midpoints across different bases, tasks, and models
• Cross-model pairwise cosine similarity of zero-shot control responses = 0.603 (n=12,720 pairs, t=35.1, p=4.3×10⁻²⁶² vs. experimental)finding0.724
  Experiment 3 comparison: zero-shot control shows lower semantic convergence than experimental condition
• Mean-difference patching in a two-layer ReLU circuit flips the decision to class-A by activating a third hidden unit that is silent for all natural class-A inputsfinding0.723
  Synthetic theoretical example showing pernicious divergence via hidden pathway activation
• In simulations, positive evidence threshold for Bayesian model reduction corresponds to ΔF ≤ −3, equivalent to odds ratio of exp(−3) ≈ 0.05 (reduced model ~20 times more likely than full model).finding0.723
  Quantitative threshold used for accepting reduced models; linked to Bayes factor of ~20
• A minimal prompt change can flip behavior.quote0.718
  Illustrates sensitivity to anchors.
• Using only loss-scale balancing (log transformation) yields Δp = +0.06±0.09 on NYUv2.finding0.717
  Ablation study component effectiveness.