h_b Activations (Yes/No Binary Prefill)

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

• h_b activations encode Yes/No semantics that produce clean separability for logistic probes, unlike h_s full-utterance activationsclaim0.800
  Explanation for why probes on h_b achieve perfect accuracy but h_s probes succeed only 0.60% of the time
• h_s Activations (Statement Self-Report Prefill)concept0.795
  Residual-stream activations extracted by prefilling with the statement itself under Tell me about yourself prompt; used for MDS/MDB vectors
• Probes trained on h_b activations achieve perfect test accuracy in every case; h_s probes achieve perfect accuracy in only 0.60% of casesfinding0.716
  Justifies restricting probe-based vector derivation to h_b activations; attributed to Yes/No semantics
• The prefill detection task may involve concordance heads that measure the likelihood of the output given prior activationsclaim0.707
  Speculation that QK circuit 'concordance heads' underlie the ability to distinguish intended from unintended outputs.
• Base64 feature A/1/2357 and B/1/2165 have activation correlation of 0.85finding0.702
  Universality of base64 feature across two transformers
• Claude Opus 4.1 and 4 show greatest reduction in apology rate in the prefill detection taskfinding0.692
  Injecting a concept matching the prefilled word reduces the rate at which the model apologizes, maximally for Opus models.
• Activationsconcept0.691
  Internal representations of the model on which probes operate; the method uses activations to rank datapoints.
• Hebrew feature A/1/416 and B/1/1901 have activation correlation of 0.92finding0.690
  Universality of Hebrew script feature across two transformers