method
active
method:linear-and-logistic-probesLinear and logistic probes
Used to characterize counting representations.
Neighborhood — ranked by edge-count
Papers (1)
paper
Methods (2)
method
- Linear Proberelated_toSimple linear classifiers trained on model activations used as the probing technique within the introduced method.
- Linear Probingrelated_toUsed to evaluate representation quality across VTAB tasks
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.
- Classifiers used for token filtering and reward shaping.
- Nguyen et al. trained linear probes on activations to distinguish evaluation from deployment scenarios.
- Standard linear probing technique; compared to mass-mean probing for classification accuracy and causal implication
- Method for fitting a linear classifier on collected activations to predict task-relevant features
- Logistic regression trained on GSM8k training set to predict answer correctness from projection features along reflection direction
- Uses normalized probe weights as steering vector.
- Interpretation of Fig. 9.
- Method to select d_steer top-activated SAE features for constructing control vectors