Linear probe achieves 100% classification accuracy for almost all components in Pythia-6.9B gender task

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

• Probe achieves selectivity of 4.20 on pythia-410m, slightly exceeding DAS selectivity of 3.96finding0.808
  Key result showing that for models larger than pythia-70m, probe selectivity matches or exceeds DAS selectivity
• Pythia-6.9B achieves 100% accuracy on gendered pronoun prediction taskfinding0.808
  Baseline result confirming the model has fully learned the gender prediction task before probing
• pythia-14m achieves only 0.38 accuracy on npi_ever_subj-relc taskfinding0.793
  Baseline accuracy showing small models fail on harder NPI licensing tasks
• Linear Probe for Evaluation Awarenessmethod0.785
  Nguyen et al. trained linear probes on activations to distinguish evaluation from deployment scenarios.
• Smaller fully trained Pythia models (31M, 70M) show slightly reduced alignment accuracy compared to larger models despite non-linear mapsfinding0.779
  Attributed to model anisotropy from saturation making hidden states harder to access
• MM probe trained on likely dataset achieves NIE of 0.70 (false→true) on LLaMA-2-13B, surprisingly strong but weaker than truth probesfinding0.778
  Likely-trained MM probe is a surprisingly effective causal baseline due to correlation between truth and probability on sp_en_trans
• Probes trained on h_b activations achieve perfect test accuracy in every case; h_s probes achieve perfect accuracy in only 0.60% of casesfinding0.772
  Justifies restricting probe-based vector derivation to h_b activations; attributed to Yes/No semantics
• Linear alignment map ϕ_lin IIA tracks DNN accuracy during Pythia-410m training progression on IOI taskfinding0.760
  Suggests linear maps may be better correlated with genuine task implementation than non-linear maps