Calibration (Model Calibration)
Degree to which predicted probabilities reflect true outcome frequencies; improved with temperature scaling.
Degree to which predicted probabilities reflect true outcome frequencies; improved with temperature scaling.
Prompting technique that encourages step‑by‑step reasoning traces from language models.
When a model forgets previously learned tasks while being trained on new tasks in sequence.
Loss function for multi‑class classification comparing predicted distributions to one‑hot targets.
Neural network architecture using convolutions, effective for images, audio, and spatial data.
Attestation scheme to show a custodian holds sufficient assets, ideally with liabilities proofs and user verification.
Liquidity pool on Curve for swapping correlated assets, often with gauges and incentives.
Security achieved by aligning incentives and penalties so that attacks are economically irrational.
Study of secure communication techniques, including hashing, signatures, and encryption.