AI Paper Atlas

Research map / Label-efficient and robust learning / Recognition and long-tail

Soft Calibration Objectives for Neural Networks

Archit Karandikar, Nicholas Cain, Dustin Tran, Balaji Lakshminarayanan, Jonathon Shlens, Michael Curtis Mozer, Rebecca Roelofs

NeurIPS 2021 · Label-efficient and robust learning › Recognition and long-tail · approach: Confidence calibration objectives · 8 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 S-AvUC secondary loss (ablation)Focal · CIFAR-100ECE (%): 4.66 -> 1.57Improved
add SB-ECE secondary loss (ablation)Focal · CIFAR-100ECE (%): 4.66 -> 2.30Improved
add SB-ECE secondary loss (ablation)Focal · CIFAR-10ECE (%): 5.45 -> 1.19Improved
add S-AvUC secondary loss (ablation)Focal · CIFAR-10ECE (%): 5.45 -> 1.58Improved
add S-AvUC secondary loss (ablation)NLL · ImageNetECE (%): 3.81 -> 2.26Improved
add SB-ECE secondary loss (ablation)NLL · ImageNetECE (%): 3.81 -> 3.11Improved
add MMCE secondary loss (ablation)NLL · CIFAR-100ECE (%): 9.10 -> 4.77Improved
add S-AvUC to Focal costs accuracy (ablation)Focal · CIFAR-10accuracy (%): 95.0 -> 94.4Hurt
add post-hoc temperature scaling (ablation)Focal · CIFAR-100ECE (%): 4.66 -> 6.47Hurt
add post-hoc temperature scaling (ablation)Focal + S-AvUC · CIFAR-100ECE (%): 1.57 -> 4.15Hurt
add post-hoc temperature scaling (ablation)NLL · CIFAR-100ECE (%): 9.10 -> 5.36Improved
add MMCE to NLL, then temperature scaling (ablation)NLL + MMCE · ImageNetECE (%): 3.12 -> 2.18No 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