DAS on randomly initialized small networks (|N|=16) achieves only 0.50 IIA (chance), cannot construct new behaviors

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

• DAS on oversized randomly initialized network (|N|=4096 for 16-dim input) achieves 0.64 IIA by searching random structurefinding0.883
  Shows that overly large hidden dimensions allow DAS to find random causal structures; calibration check.
• DAS achieves 100% IIA on hierarchical equality task with |N|=16, intervention size 8, Layer 1finding0.759
  DAS discovers a perfect alignment between the feed-forward network and the Both Equality Relations high-level model.
• DAS behavioral loss achieves IIA of 0.997±0.001 on synthetic 10-class dataset training/test setsfinding0.758
  IIA baseline for DAS behavioral loss on synthetic dataset
• EI of ER random networks converges to -log2(p) with increasing size, with a phase transition at average degree ≈ log2(N).finding0.757
  From Klein & Hoel (2020) analysis of artificial complex networks.
• DAS learning rate of 5e-3 outperforms 1e-3 (used in Wu et al. 2023) for small training sets in CausalGymfinding0.748
  Hyperparameter tuning result for DAS; different from prior work due to smaller training set size
• DAS overcomes the localist limitation of prior causal abstraction by allowing individual neurons to play multiple roles via non-standard basesclaim0.745
  Central claim motivating DAS over prior methods.
• 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.finding0.737
  Table 2, row 3, showing equivalence when prior preferences match rewards.
• Several Mixtral-8x7B samples could not be initialized as valid networks using PyPhi under IIT 4.0 and were excluded.finding0.734
  Methodological limitation disproportionately affecting the largest MoE model, constraining generalizability.