Zero-Shot Generalization

Ability to predict correctly for stimulus-action pairs never previously experienced by inferring structural rules; key measure for TEM-t performance.

Neighborhood — ranked by edge-count

paper

concept

Abstract Structural Knowledge
associated_with
Knowledge of spatial rules (e.g., North+East+South+West=0) that generalises across environments and enables zero-shot prediction.

cosine ≥ 0.65 · no typed edge

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

Generalizationconcept0.812
Ability to apply learned solutions to novel circumstances.
Generalisationconcept0.799
Ability to respond appropriately to novel situations based on past regularities; fundamental to learning and intelligence.
zero-shot predictionconcept0.794
Prediction without task-specific training; Evee achieves 0.991 AUROC on indels in zero-shot mode.
Zero-Shot Control Conditionconcept0.788
Control omitting any induction and presenting only the final experiential query
generalization (abstraction)concept0.780
Abstracting from specific memories (e.g., specific leaves) to general lessons (food).
scope generalizationconcept0.763
Generalization from 2-digit to 3-4 digit arithmetic; limited by mismatch dr.
Few-shot learningconcept0.759
Test-time adaptation from a small number of examples without parameter updates.
Zero-Shot Model Stitchingconcept0.756
Model stitching without learning a stitching layer, demonstrating strong alignment across different model training regimes