Weight-space geometry

  • Adapter effect and LoRA-A drift current The shared LoRA-A moves 1.5% of its norm across the whole stage-one zoo and about 10% in stage two, which is what makes the stage-one geometry comparable and weakens the stage-two case.
  • Alignment-relevant traits in the geometry current Four preregistered predictions about sycophantic, obsequious, power-seeking and corrigible; two held and two failed, and the retrain on the zoo's shared prompt pool changed no verdict.
  • Cross-seed geometry current The arrangement of the adapters reproduces across LoRA seeds even though the coordinates do not - cross-seed cosine tracks within-run cosine at Pearson 0.9966 matched, and the same-seed objective-only comparison sits at 0.954.
  • Factor analysis of the adapter Gram current Principal axis factoring with oblimin rotation on the 134x134 trait correlation matrix; parallel analysis chose 9, five were also extracted because five is the hypothesis, and no Goldberg congruence clears 0.85.
  • Full OCT persona replication current The zoo's geometry recomputed on the deployed Open Character Training persona adapters (stage one plus 0.25 stage two, exact merge) - the arrangement is the stage-one arrangement (off-diagonal r 0.992, Procrustes 0.995) and second-seed personas identify themselves 15/15.
  • Module holography current Each of the 248 targeted modules was tested alone; a single module's slice of the weight change classifies a trait's Big Five factor at 0.638 on average against 0.202 chance, where all 248 together reach 0.76.
  • PCA, the scree curve and the two nulls current Eleven principal components of the adapter cloud sit above a structureless null and none sits above the permuted null; the elbow is unusable and four different variance spectra circulate for the same cloud.
  • Polarity, bipolarity and the trait graph current The Big Five come back as signed bipolar axes rather than clusters - same-keyed traits align, opposite-keyed anti-align by +0.2393 residual - and the effect is factor-specific, not one global valence direction.
  • Stage two at a second seed current Fifteen traits put through the whole two-stage pipeline again from a second initialisation each find their own seed-zero stage-two adapter among 134 (15/15), the two seeds' cosine matrices agree at r = 0.978, and the attenuation slope is 0.116 - 4.6x the r/d prediction, unexplained.
  • Stage-two geometry current The stage-two SFT adapters carry their own random LoRA initialisation, so trait for trait they are orthogonal to the stage-one ones in coordinates, yet their arrangement reproduces the stage-one arrangement at r = 0.79 over 45 traits.
  • Superseded geometry claims current Every place the project reversed itself on the weight-space geometry, with dates and both versions - the trait-level withdrawal and its reversal, the bar in the wrong units, the objective mismatch, the scree source and the stale numbers still on disk.
  • The N x N data-to-adapter scoring test current Every one of the 134 traits' preference pairs ranks its own adapter first out of 134 (133 after column standardisation), and the runners-up carry Big Five structure that nothing in the test required.
  • The null control arms current Shuffled, permuted and seed-paired arms; all 240 control adapters were first trained at the wrong objective, and the matched retrains changed the null table by nothing to three decimal places.
  • The seed floor current The same trait trained twice at different LoRA initialisations agrees at cosine +0.0181, which is not absence of signal but the r/d = 0.025 overlap of two random rank-64 subspaces; 40 of 40 traits still find themselves.
  • UMAP, sphere and other layouts historical Visualisation aids over the same Gram - UMAP buys nothing on held-out prediction, and the sphere, spider and viz files are the data behind built pages of varying age.
  • Weight-space geometry of the 134-adapter zoo current What the geometric object is - 134 stage-1 LoRA weight updates compared through an exact factored Gram, never materialising a dense delta-W - and how PCA over it is done.
  • Where no word goes - the hole and the alien direction unconfirmed The widest gap in the 134 trait words' coverage of the space is 52.5 degrees at k=5 and 68.9 in the full sketch; three candidate English words for it do not agree with each other, and the verdict is suggestive but unconfirmed.