Calibrated Few-Shot Prompting

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

• Few-shot learningconcept0.819
  Test-time adaptation from a small number of examples without parameter updates.
• k-shot promptingmethod0.796
  Prompting technique where k example pairs are provided as anchors.
• few-shot threshold (k50/θ50)concept0.782
  Shot count needed to reach 50% accuracy; reflects when anchoring strength crosses critical value.
• in-context k-shot promptingmethod0.777
  Use k examples as anchors with no parameter update.
• Language models are few-shot learners (Brown et al., 2020)concept0.744
  Demonstrated transformers on mathematical understanding and logic; cited to motivate transformer versatility.
• Few-shot linear probe steering baselinemethod0.744
  Constructing steering vectors from the difference of mean activations on positive and negative examples, for comparison.
• zero-shot predictionconcept0.725
  Prediction without task-specific training; Evee achieves 0.991 AUROC on indels in zero-shot mode.
• The ordering of few-shot thresholds k50 and transition widths aligns with k50 ∝ dr/ρd.claim0.724
  Interpretation of E2 results.