DAS learning rate of 5e-3 outperforms 1e-3 (used in Wu et al. 2023) for small training sets in CausalGym

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

• DAS achieves overall odds-ratio of 10.24 on pythia-410m averaged across all CausalGym tasksfinding0.790
  Numerical result for pythia-410m
• DAS finds causal effect at all training timesteps including when model is just initialisedfinding0.787
  Corroborates Wu et al. 2023 finding that DAS expressivity inflates causal effect estimates
• DAS consistently finds the most causally-efficacious features across all pythia model sizes in CausalGymfinding0.763
  Main benchmark result showing DAS superiority over probing, diff-in-means, PCA, k-means, LDA, and random
• Trainable intervention (DAS) finds sparser gender representations than linear probing, suggesting probing overestimates causal coverageclaim0.762
  Interpretive claim from Case Study II about the distinction between correlational probes and causal interventions
• Task accuracy on CausalGym increases consistently with model scale from 0.62 (14M) to 0.89 (6.9B)finding0.759
  Scaling result showing larger pythia models perform better on CausalGym linguistic tasks
• Under reward shaping (G=100, H=-100, F=0), Active Inference scored 99.52, Bayesian RL 99.77, Q-learning 95.56, with nearly identical behavior between belief-based agents.finding0.751
  Table 2, row 3, showing equivalence when prior preferences match rewards.
• DAS behavioral loss achieves IIA of 0.997±0.001 on synthetic 10-class dataset training/test setsfinding0.750
  IIA baseline for DAS behavioral loss on synthetic dataset
• DAS achieves substantial causal effect even on arbitrary input-output mappings where no causal mechanism should existfinding0.749
  Replication of Wu et al. 2023 finding; DAS expressivity concern validated in CausalGym setup