Language Model Calibration on SCITLDR (dev) benchmark leaderboard · SOTA2 Research

Language Model Calibration on SCITLDR (dev)

0.072smECEACUTE cosine Platt

Evaluation Results

Method	Links
ACUTE cosine Platt2026.06		0.072	0.678	0.721	0.925	0.774
ACUTE pca20 Isotonic2026.06		0.072	0.692	0.724	0.925	0.78
ACUTE pca20 Beta2026.06		0.072	0.699	0.723	0.925	0.782
ACUTE cosine Beta2026.06		0.073	0.678	0.721	0.925	0.774
ACUTE pca20 Platt2026.06		0.073	0.698	0.723	0.925	0.782
ACUTE pca10 Platt2026.06		0.074	0.694	0.722	0.925	0.779
ACUTE pca10 Beta2026.06		0.074	0.694	0.722	0.925	0.779
ACUTE pca10 Isotonic2026.06		0.075	0.686	0.723	0.925	0.777
ACUTE cosine Isotonic2026.06		0.078	0.673	0.718	0.924	0.771
Platt2026.06		0.079	0.668	0.721	0.925	0.771
Isotonic2026.06		0.079	0.669	0.721	0.925	0.771
Beta2026.06		0.079	0.668	0.721	0.925	0.771
HRE2026.06		0.081	0.665	0.721	0.925	0.769
NWKR2026.06		0.082	0.667	0.721	0.925	0.77
ACUTE mid act2026.06		0.084	0.689	0.722	0.925	0.778
ACUTE late act2026.06		0.084	0.681	0.722	0.925	0.775
ACUTE cosine2026.06		0.085	0.669	0.717	0.925	0.769
ACUTE early act2026.06		0.086	0.686	0.722	0.925	0.777
ACUTE pca102026.06		0.089	0.685	0.723	0.925	0.777
ACUTE pca202026.06		0.091	0.688	0.723	0.925	0.778
Raw Conf2026.06		0.152	0.366	0.721	0.925	0.667