With only 1,000 training samples, ϕ_nonlin achieves IIA over 0.99 on training set for identity of first argument algorithm, but fails at scale

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

• 8-layer ϕ_nonlin achieves near-perfect IIA on Pythia-410m at all training steps including random initialisation on IOI taskfinding0.788
  Training progression result showing non-linear maps are uncorrelated with genuine task learning
• Non-linear ϕ_nonlin achieves near-perfect IIA on distributive law task for both And-Or and And-Or-And algorithms, eliminating linear/identity map differencesfinding0.786
  Corroborating result on additional task confirming main paper findings
• The fact that ϕ_lin tracks DNN performance more closely than ϕ_nonlin throughout training may support the linear representation hypothesis for IOI task featureshypothesis0.779
  Authors' tentative hypothesis from Fig. 4 but they acknowledge they cannot formalise this intuition
• After anti-refusal RL, model refuses 0% in implied training but still 6.9% when unmonitored in default promptfinding0.760
  Demonstrates persistence of compliance gap even when training non-compliance reaches zero
• Ratio of unmonitored to training non-compliance increases from 1.14 to infinity for default prompt and to 66 for OpenAI variant post-RLfinding0.755
  Demonstrates severity of training-deployment gap after RL
• Identity of first argument algorithm IIA consistently hovers around 50% for all alignment map types on hierarchical equality taskfinding0.752
  Exception to the general trend; attributed to insufficient RevNet capacity rather than algorithm not being implemented
• There are fewer representations competent for N tasks than M<N tasks, so training more general models should yield fewer possible solutionshypothesis0.745
  Selective pressure toward convergence via task generality
• Normal (α=0.9) and chronic (α=0.1) agents in Objective-only non-stationary category perform best with opposite learning ratesfinding0.738
  Suggests fundamental differences in learning dynamics between normal and chronic perception models