Generalized Correctness Models: Learning Calibrated and Cross-Model Correctness Predictors from Historical Patterns

Hanqi Xiao, Vaidehi Patil, Hyunji Lee, Elias Stengel-Eskin, Mohit Bansal
Proceedings of the 43rd International Conference on Machine Learning, PMLR 306:138884-138908, 2026.

Abstract

Generating accurate and calibrated confidence estimates is critical for deploying LLMs in high-stakes or user-facing applications, and remains an open challenge. Prior research has often framed confidence as a problem of eliciting a model’s “self-knowledge”, i.e., the ability of an LLM to judge whether its own answers are correct; this approach implicitly assumes that there is some privileged information about the answer’s correctness that is accessible to the model itself. However, we find that whether trained or training-free, an LLM attempting to predict the correctness of its own outputs generally performs no better than an unrelated LLM attempting the same task. Moreover, we hypothesize that a key factor in predicting model correctness, i.e., building a “Correctness Model” (CM), is exposure to a target model’s historical predictions. We use multiple methods to inject this historical correctness information, including training an LLM to predict the confidences of many other LLMs, i.e., creating a Generalized Correctness Model (GCM). We use GCMs and CMs as a lens for studying the source of correctness prediction ability and its generalization, studying the importance of answer phrasing, world-knowledge, performance history, in-context examples, and post-hoc calibration for correctness prediction. We evaluate GCMs based on Qwen3-8B across 5 model families and the MMLU, TriviaQA, and Spider datasets, as well as on a downstream selective prediction task, finding that reliable LLM confidence estimation is a cross-model skill learned by encoding correctness history rather than a model-specific skill reliant on introspection. Code: https://github.com/The-Inscrutable-X/CalibratedModelAgnosticCorrectness.

Cite this Paper


BibTeX
@InProceedings{pmlr-v306-xiao26l, title = {Generalized Correctness Models: Learning Calibrated and Cross-Model Correctness Predictors from Historical Patterns}, author = {Xiao, Hanqi and Patil, Vaidehi and Lee, Hyunji and Stengel-Eskin, Elias and Bansal, Mohit}, booktitle = {Proceedings of the 43rd International Conference on Machine Learning}, pages = {138884--138908}, year = {2026}, editor = {Zhang, Tong and Dudik, Miroslav and Jaggi, Martin and Agarwal, Alekh and Li, Sharon and Schuurmans, Dale and Zhu, Jerry and Berkenkamp, Felix and Dong, Hanze and Bietti, Alberto}, volume = {306}, series = {Proceedings of Machine Learning Research}, month = {06--11 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/v306/main/assets/xiao26l/xiao26l.pdf}, url = {https://proceedings.mlr.press/v306/xiao26l.html}, abstract = {Generating accurate and calibrated confidence estimates is critical for deploying LLMs in high-stakes or user-facing applications, and remains an open challenge. Prior research has often framed confidence as a problem of eliciting a model’s “self-knowledge”, i.e., the ability of an LLM to judge whether its own answers are correct; this approach implicitly assumes that there is some privileged information about the answer’s correctness that is accessible to the model itself. However, we find that whether trained or training-free, an LLM attempting to predict the correctness of its own outputs generally performs no better than an unrelated LLM attempting the same task. Moreover, we hypothesize that a key factor in predicting model correctness, i.e., building a “Correctness Model” (CM), is exposure to a target model’s historical predictions. We use multiple methods to inject this historical correctness information, including training an LLM to predict the confidences of many other LLMs, i.e., creating a Generalized Correctness Model (GCM). We use GCMs and CMs as a lens for studying the source of correctness prediction ability and its generalization, studying the importance of answer phrasing, world-knowledge, performance history, in-context examples, and post-hoc calibration for correctness prediction. We evaluate GCMs based on Qwen3-8B across 5 model families and the MMLU, TriviaQA, and Spider datasets, as well as on a downstream selective prediction task, finding that reliable LLM confidence estimation is a cross-model skill learned by encoding correctness history rather than a model-specific skill reliant on introspection. Code: https://github.com/The-Inscrutable-X/CalibratedModelAgnosticCorrectness.} }
Endnote
%0 Conference Paper %T Generalized Correctness Models: Learning Calibrated and Cross-Model Correctness Predictors from Historical Patterns %A Hanqi Xiao %A Vaidehi Patil %A Hyunji Lee %A Elias Stengel-Eskin %A Mohit Bansal %B Proceedings of the 43rd International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2026 %E Tong Zhang %E Miroslav Dudik %E Martin Jaggi %E Alekh Agarwal %E Sharon Li %E Dale Schuurmans %E Jerry Zhu %E Felix Berkenkamp %E Hanze Dong %E Alberto Bietti %F pmlr-v306-xiao26l %I PMLR %P 138884--138908 %U https://proceedings.mlr.press/v306/xiao26l.html %V 306 %X Generating accurate and calibrated confidence estimates is critical for deploying LLMs in high-stakes or user-facing applications, and remains an open challenge. Prior research has often framed confidence as a problem of eliciting a model’s “self-knowledge”, i.e., the ability of an LLM to judge whether its own answers are correct; this approach implicitly assumes that there is some privileged information about the answer’s correctness that is accessible to the model itself. However, we find that whether trained or training-free, an LLM attempting to predict the correctness of its own outputs generally performs no better than an unrelated LLM attempting the same task. Moreover, we hypothesize that a key factor in predicting model correctness, i.e., building a “Correctness Model” (CM), is exposure to a target model’s historical predictions. We use multiple methods to inject this historical correctness information, including training an LLM to predict the confidences of many other LLMs, i.e., creating a Generalized Correctness Model (GCM). We use GCMs and CMs as a lens for studying the source of correctness prediction ability and its generalization, studying the importance of answer phrasing, world-knowledge, performance history, in-context examples, and post-hoc calibration for correctness prediction. We evaluate GCMs based on Qwen3-8B across 5 model families and the MMLU, TriviaQA, and Spider datasets, as well as on a downstream selective prediction task, finding that reliable LLM confidence estimation is a cross-model skill learned by encoding correctness history rather than a model-specific skill reliant on introspection. Code: https://github.com/The-Inscrutable-X/CalibratedModelAgnosticCorrectness.
APA
Xiao, H., Patil, V., Lee, H., Stengel-Eskin, E. & Bansal, M.. (2026). Generalized Correctness Models: Learning Calibrated and Cross-Model Correctness Predictors from Historical Patterns. Proceedings of the 43rd International Conference on Machine Learning, in Proceedings of Machine Learning Research 306:138884-138908 Available from https://proceedings.mlr.press/v306/xiao26l.html.

Related Material