method
active
method:noised-label-vit-memorization-setupNoised-label ViT memorization setup
10% random labels force exact memorization in ViT.
Neighborhood — ranked by edge-count
Papers (1)
paper
Datasets (1)
dataset
- ViT training data.
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.
- Vision disentanglement.
- Motivates noise analysis.
- Training distribution where special point p=[1,1,1,0,...] appears with fraction repeat_frac.
- Imagine instruction raises property name in J-lens while J-orthogonalized property probe stays at baseline.finding0.667J-space privileged.
- Validates behavioral probe.
- The paper's proposed training-free prompting strategy that prompts the model to verbalize a probability distribution over a set of responses rather than generating a single response
- Interpretation of illusion.
- Maximum Φ over all subsystems; represents the most integrated subsystem (main complex) under IIT 3.0.