AI Paper Atlas

Research map / Learning theory and algorithms / Prediction and decision losses

Black-Box Assisted Regression: Phase Transitions and Minimax Optimality

Yan Zhou

ICML 2026 · Learning theory and algorithms › Prediction and decision losses · approach: Distributionally robust optimization · 0 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
add holdout safe selection (fallback to black-box) to the residual estimator (ablation)Residual estimator · 1D synthetic regression (n=20, delta=0.0)MSE: 0.0335 -> 0.0027Improved
add holdout safe selection to the residual estimator in high dimensions with a poor prior (ablation)Residual estimator · High-dimensional regression (d=20, n=100, delta=1.0)MSE: 1.026 -> 1.007Improved
learn a nonlinear residual instead of a weighted ensemble of black-box and scratch (ablation)Safe Residual vs WGT · 1D synthetic regression (n=20, delta=0.8)MSE: 0.1452 -> 0.0971Improved
learn a residual around the black-box instead of training from scratch (ablation)Safe Residual vs Scratch · 1D synthetic regression (n=50, delta=0.0)MSE: 0.0325 -> 0.0006Improved
use residual learning when the black-box is badly biasedResidual estimator · 1D synthetic regression (n=20, delta=1.5)MSE: 2.2496 -> 0.1461Improved
use zero-initialized safe residual instead of validation-tuned weighted ensembleCLIP on CIFAR-100 · CIFAR-100 (n=2000 labeled)accuracy (%): 63.9 -> 66.7Improved
use zero-initialized safe residual instead of validation-tuned weighted ensembleCLIP on CIFAR-100 · CIFAR-100 (n=200 labeled)accuracy (%): 54.5 -> 60.4Improved
use safe residual instead of weighted ensemble in ultra-low-sample NLP regimeQwen3-8B on AG News · AG News (n=64 labeled)accuracy (%): 54.7 -> 74.7Improved
use safe residual instead of weighted ensemble with more labeled dataQwen3-8B on AG News · AG News (n=256 labeled)accuracy (%): 75.6 -> 77.0Improved
concatenate black-box output to features instead of residual learningCLIP on CIFAR-100 · CIFAR-100 (n=100 labeled)accuracy (%): 0.4 collapse versus 59.8 for safe residualHurt
add zero-initialization of residual head (ablation)Safe Residual neural head · CIFAR-100 and AG Newsaccuracy / early-training varianceImproved
vary validation fraction rho from 0.15 to 0.3 (ablation)Safe Residual · CIFAR-100accuracyNo clear effect

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy