Trainable intervention (DAS) finds sparser gender representations than linear probing, suggesting probing overestimates causal coverage

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

• DAS finds causal effect at all training timesteps including when model is just initialisedfinding0.807
  Corroborates Wu et al. 2023 finding that DAS expressivity inflates causal effect estimates
• Probe-based data attribution effectively reduces harmful behaviors via data interventionsclaim0.786
  Authors' central interpretive assertion that their method meaningfully mitigates unwanted behaviors.
• DAS's access to model outputs during training is responsible for much of its advantage over other interpretability methodsclaim0.776
  Author interpretation of selectivity results showing DAS advantage diminishes when controlling for expressivity
• Direct probes over learned activations in standard basis may fail to reveal the actual causal role of representations because they are highly distributedclaim0.771
  Supported by the finding that non-trivial rotations are required to find aligned representations.
• DAS achieves 100% IIA on hierarchical equality task with |N|=16, intervention size 8, Layer 1finding0.769
  DAS discovers a perfect alignment between the feed-forward network and the Both Equality Relations high-level model.
• Probe-based data attribution bridges interpretability work (probes/activations) with data-centric alignment.claim0.769
  Conceptual framing: integrates mechanistic interpretability tools with alignment-focused data curation.
• DAS achieves substantial causal effect even on arbitrary input-output mappings where no causal mechanism should existfinding0.769
  Replication of Wu et al. 2023 finding; DAS expressivity concern validated in CausalGym setup
• Simple difference-in-mean probes generalize as well as other probing techniques while identifying directions which are more causally implicated in model outputsclaim0.769
  Key methodological claim: MM probes are both competitive in accuracy and superior in causal influence