method
active
method:localization-probeLocalization probe
Causal transformer probe predicting whether a token is in the same Entity as its predecessor.
Neighborhood — ranked by edge-count
Papers (1)
paper
Frameworks (1)
framework
- Core pipeline turning probe-read features into scalable reward functions for open-ended tasks.
Methods (2)
method
- Classification proberelated_toAttention probe predicting whether an Entity is hallucinated; AUC .94 on validation.
- Two-step localize-then-classify pipeline for hallucination monitoring and reward.
Related by similarity (8)
cosine ≥ 0.65 · no typed edgeEntities in the same semantic neighborhood but without a typed relation to this one — candidates for new edges or unrecognized duplicates.
- The ability of probes trained on one dataset to transfer accurately to topically and structurally different datasets
- Interpretability tools that decode information from internal model activations; here, linear probes are used for data attribution.
- Standard interpretability technique used to read feature beliefs.
- Chinn et al. showed that tactile target experience promotes earlier mirror self-recognition in infants; noted as a future extension
- Reward probe grading correction interventions (5-class).
- Linear classifier approach applied to model activations to identify which training datapoints caused undesired behaviors in post-training.
- Internals-based approach to which the logit monitor is complementary.
- Earlier interpretability method applying classifiers to DNN hidden representations; shares complexity-accuracy dilemma with causal abstraction