Few-shot learning

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

• few-shot promptingmethod0.860
  Providing k labeled examples in the prompt to steer model behavior.
• Calibrated Few-Shot Promptingmethod0.819
  Baseline method: sweeps over shot count and resamples prompts; calibrates threshold for P(TRUE)-P(FALSE); performed surprisingly weakly
• Language models are few-shot learners (Brown et al., 2020)concept0.798
  Demonstrated transformers on mathematical understanding and logic; cited to motivate transformer versatility.
• Many-Shot Promptingmethod0.785
  Technique using 0-20 in-context examples exhibiting a target trait to elicit behavioral shifts, used to validate persona vector monitoring
• few-shot threshold (k50/θ50)concept0.771
  Shot count needed to reach 50% accuracy; reflects when anchoring strength crosses critical value.
• Many-Shot Jailbreakingconcept0.763
  Phenomenon where providing many untruthful or undesirable examples in context steers LM behaviour, related to reflective traits.
• Few-shot linear probe steering baselinemethod0.761
  Constructing steering vectors from the difference of mean activations on positive and negative examples, for comparison.
• Zero-Shot Generalizationconcept0.759
  Ability to predict correctly for stimulus-action pairs never previously experienced by inferring structural rules; key measure for TEM-t performance.