SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2605.30447 · ICML · 2026

Calibrated Preference Learning: The Case of Label Ranking

Sebastian Vollmer, Eyke Üllermeier, Timo Kaufmann, Viktor Bengs, Santo Thies

arXiv · PDF · Open in the Atlas

Code that ran

We lifted 10 functions out of this paper's own repositories and ran 8 of them in a sandbox. "Ran" means the function executed on a synthesized input and returned a value. It is not a reproduction of the paper's results.

RepositoryRoleRan
Advueu963/Calibrated_Preference_Learning canonical 8 of 10
FunctionStatusWhere it lives
build_plackett_luce_mlp Ran Advueu963/Calibrated_Preference_Learning/src/cal_pref/preference_models.py
pointer only (licence: NONE) · get_code("728abea8736269ca")
build_preference_mlp Ran Advueu963/Calibrated_Preference_Learning/src/cal_pref/preference_models.py
pointer only (licence: NONE) · get_code("7bae38437e4708c1")
find_model_file Ran Advueu963/Calibrated_Preference_Learning/analyze_rb_logit_diff.py
pointer only (licence: NONE) · get_code("d4d57f528c187077")
is_label_ranker Ran Advueu963/Calibrated_Preference_Learning/scikit-lr/sklr/base.py
pointer only (licence: NONE) · get_code("765300dd23f18fd4")
is_partial_label_ranker Ran Advueu963/Calibrated_Preference_Learning/scikit-lr/sklr/base.py
pointer only (licence: NONE) · get_code("140075aae413f8ba")
kendal_distance Ran Advueu963/Calibrated_Preference_Learning/src/cal_pref/utils.py
pointer only (licence: NONE) · get_code("9f6164abd81077de")
load_lr_data Ran Advueu963/Calibrated_Preference_Learning/src/cal_pref/utils.py
pointer only (licence: NONE) · get_code("876657632e892fea")
synthetic_data Ran Advueu963/Calibrated_Preference_Learning/src/cal_pref/utils.py
pointer only (licence: NONE) · get_code("67bf10f432d3af10")
load_scores Not yet run Advueu963/Calibrated_Preference_Learning/analyze_rb_logit_diff.py
pointer only (licence: NONE) · get_code("e83ea18ca7066d9e")
parse_leaderboard Not yet run Advueu963/Calibrated_Preference_Learning/analyze_rb_logit_diff.py
pointer only (licence: NONE) · get_code("4355a3a21edbb8ac")

Repositories linked to this paper

Some links come from the archived Papers with Code dataset (CC BY-SA 4.0): attribution and licence.

Abstract

Calibration, the alignment of predicted probabilities with true outcome frequencies, is essential for reliable decision-making. While extensively studied for classification and regression, calibration has not been formally addressed for probabilistic label ranking, where the goal is to predict a distribution over orderings of a label set. Naively treating rankings as classes ignores their structure and fails to capture important modalities such as pairwise and top-k predictions. We formalize calibration for label ranking and develop a hierarchy of notions covering full rankings, sub-rankings, and top-k rankings. We prove that full-rank calibration implies the others but not conversely, and sub-ranking and top-k calibration are incomparable. Empirically, we find popular label ranking models are often poorly calibrated, with substantial differences between sub-ranking and top-k metrics. Applying our framework to RLHF reward models, we find that calibration correlates strongly but not perfectly with benchmark accuracy, suggesting it captures a meaningful quality dimension beyond top-1 accuracy. These findings motivate future work on understanding the downstream effects of miscalibration and developing methods to correct it.

For agents

The same record, over MCP at https://syntology.ai/mcp:

get_harvested_code_for_paper("2605.30447")
get_code_for_paper("2605.30447")
have("2605.30447")

Connect an agent — have() is free.