DAS achieves overall odds-ratio of 10.24 on pythia-410m averaged across all CausalGym tasks

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

• DAS consistently finds the most causally-efficacious features across all pythia model sizes in CausalGymfinding0.833
  Main benchmark result showing DAS superiority over probing, diff-in-means, PCA, k-means, LDA, and random
• DAS learning rate of 5e-3 outperforms 1e-3 (used in Wu et al. 2023) for small training sets in CausalGymfinding0.790
  Hyperparameter tuning result for DAS; different from prior work due to smaller training set size
• LDA barely outperforms random features across all pythia model sizes in CausalGymfinding0.783
  Surprising negative result for LDA despite being a supervised method
• Task accuracy on CausalGym increases consistently with model scale from 0.62 (14M) to 0.89 (6.9B)finding0.760
  Scaling result showing larger pythia models perform better on CausalGym linguistic tasks
• Probe achieves selectivity of 4.20 on pythia-410m, slightly exceeding DAS selectivity of 3.96finding0.759
  Key result showing that for models larger than pythia-70m, probe selectivity matches or exceeds DAS selectivity
• DAS runs in 502 seconds for hierarchical equality vs. estimated 6e8 seconds for exhaustive brute-force searchfinding0.753
  DAS runtime is invariant with number of testing hypotheses, unlike brute-force search.
• MDS achieves global win proportion of 89.5% on SJTs across 14 LLMs and four injection stridesfinding0.751
  MDS dominates in open-ended generation by global win proportion metric (Table 2)
• pythia-14m achieves only 0.38 accuracy on npi_ever_subj-relc taskfinding0.742
  Baseline accuracy showing small models fail on harder NPI licensing tasks