DAS finds better alignments than brute-force search by using gradient descent rather than exhaustive discrete search

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

• Distributed Alignment Search (DAS)framework0.834
  Practical method by Geiger et al. for finding distributed causal abstractions using gradient descent
• Brute-Force Alignment Searchmethod0.785
  Baseline method that exhaustively searches discrete spaces of localist alignments between high-level variables and neuron groups.
• We hypothesized that divergence could influence IIA when transferring the DAS alignment to OOD settingshypothesis0.760
  Motivating hypothesis for the OOD experiment testing practical utility of divergence reduction
• Larger hidden representations create more random structure that DAS can search through, allowing manipulation of counterfactual behavior even in randomly initialized networkshypothesis0.754
  Tested in Section 4.4 calibration experiment; confirmed by findings.
• For DNA database search, it is usually more efficient to run many sequential searches in parallel than to parallelize individual searches.claim0.754
  Observation from practical experience with DNA sequence comparison.
• DAS achieves 100% IIA on hierarchical equality task with |N|=16, intervention size 8, Layer 1finding0.753
  DAS discovers a perfect alignment between the feed-forward network and the Both Equality Relations high-level model.
• Probe-based ranking outperforms gradient-based and LLM-judge methods at lower costclaim0.749
  Authors' claim that their approach is both more effective in reduction and cheaper than prior methods.
• DAS achieves substantial causal effect even on arbitrary input-output mappings where no causal mechanism should existfinding0.746
  Replication of Wu et al. 2023 finding; DAS expressivity concern validated in CausalGym setup