Papers with calibration methods

21 papers
Bag of Tricks for In-Distribution Calibration of Pretrained Transformers (2023.findings-eacl)

Copied to clipboard

Challenge: Recent studies show that pre-trained language models (PLMs) often predict over-confidently.
Approach: They propose to use ensemble learning and data augmentation to improve confidence calibration for PLMs by combining calibration techniques with a trade-off between accuracy and classification.
Outcome: The proposed calibration method improves classification accuracy and confidence in pre-trained language models by combining several calibration techniques.
A Close Look into the Calibration of Pre-trained Language Models (2023.acl-long)

Copied to clipboard

Challenge: Pre-trained language models (PLMs) may fail in giving reliable estimates of their predictive uncertainty.
Approach: They conduct fine-grained control experiments to study the dynamic change in PLMs’ calibration performance in training.
Outcome: The proposed methods significantly reduce PLMs’ confidence in wrong predictions.
Reconfidencing LLMs from the Grouping Loss Perspective (2024.findings-emnlp)

Copied to clipboard

Challenge: Existing methods to calibrate confidence scores for large language models often overlook biases towards certain groups, such as specific nationalities.
Approach: They propose a method to calibrate confidence scores of Large Language Models by considering different groups, a process they call reconfidencing.
Outcome: The proposed method mitigates biases against minority groups, the authors show . they show that the proposed method is more reliable than existing methods .
Calibrated Language Model Fine-Tuning for In- and Out-of-Distribution Data (2020.emnlp-main)

Copied to clipboard

Challenge: Pre-trained language models suffer from severe miscalibration for both in-distribution and out-of-difference data due to over-parameterization.
Approach: They propose a regularized method to improve in-distribution and out-of-distance calibrations by using on-manifold regularization and off-manfold regularisation.
Outcome: The proposed method outperforms existing methods for text classification in terms of expectation calibration error, misclassification detection, and OOD detection on six datasets.
Generative Calibration for In-context Learning (2023.findings-emnlp)

Copied to clipboard

Challenge: In-context learning is one of the most exciting features of large language models . performance is sensitive to various configurations of the prompt, such as the choice or order of the training examples.
Approach: They propose to calibrate the in-context predictive distribution by adjusting the label marginal . they find that the proposed method outperforms the ICL and state-of-the-art calibration methods .
Outcome: The proposed method outperforms state-of-the-art methods by 27% absolute in macro-F1.
VSCBench: Bridging the Gap in Vision-Language Model Safety Calibration (2025.findings-acl)

Copied to clipboard

Challenge: Existing safety calibration methods focus on model undersafety, where the model responds to hazardous queries, while neglecting oversafetiness, where models refuse to answer safe queries.
Approach: They propose safety calibration which addresses both undersafety and oversafetiness by comparing model responses to a novel dataset of 3,600 image-text pairs.
Outcome: The proposed methods have been used to evaluate safety calibration across image-centric and text-centric scenarios.
Extracting Chemical-Protein Interactions via Calibrated Deep Neural Network and Self-training (2020.findings-emnlp)

Copied to clipboard

Challenge: Several natural language processing methods have been used to extract interactions between chemicals and proteins from biomedical text data.
Approach: They propose a method to extract chemical–protein interactions from biomedical text data . they use a pre-trained language-understanding model and calibration techniques to estimate uncertainty .
Outcome: The proposed approach achieves state-of-the-art performance on the Biocreative VI ChemProt task while preserving higher calibration abilities.
Re-Examining Calibration: The Case of Question Answering (2022.findings-emnlp)

Copied to clipboard

Challenge: Existing calibration methods do not provide significant gains in accuracy.
Approach: They propose a new calibration metric that better captures whether the model assigns low confidence to wrong predictions and high confidence to correct predictions.
Outcome: The proposed calibration method better captures whether the model assigns low confidence to wrong predictions and high confidence to correct predictions.
Selectively Answering Visual Questions (2024.findings-acl)

Copied to clipboard

Challenge: Large multi-modal models (LMMs) are capable of visual question answering (VQA) with unprecedented accuracy.
Approach: They propose a calibration score that can be used to quantify uncertainty in visual question answering models.
Outcome: The proposed calibration score is better calibrated than in text-only models for in-context learning.
Bayesian Calibration of Win Rate Estimation with LLM Evaluators (2024.emnlp-main)

Copied to clipboard

Challenge: Recent advances in large language models (LLMs) show the potential of using LLMs as evaluators for text quality evaluation.
Approach: They propose two methods to improve the accuracy of LLM evaluators by Bayesian inference.
Outcome: The proposed methods improve the accuracy of the win rate estimation using LLMs . the proposed methods are based on six datasets covering story generation, summarization, and instruction following tasks .
On the Calibration of Massively Multilingual Language Models (2022.emnlp-main)

Copied to clipboard

Challenge: Massively Multilingual Language Models (MMLMs) have gained popularity due to their effectiveness in cross-lingual transfer.
Approach: They investigate how well calibrated MMLMs are with respect to confidence . they find that calibration methods like temperature scaling and label smoothing improve calibration .
Outcome: The proposed models are able to generalize in languages unseen during fine-tuning, but they are not reliable across languages.
NeuroSym-Cal: Bridging the Reasoning-Execution Gap in Code Generation via Hierarchical Calibration (2026.findings-acl)

Copied to clipboard

Challenge: Existing calibration methods rely on the assumption that consensus implies correctness . Existing methods fail under systematic errors, leading to miscalibrated high-confidence predictions.
Approach: They propose a hierarchical calibration framework that measures confidence at two levels . they propose sensitivity analysis to measure local curvature of deductive process .
Outcome: The proposed framework de-saturates overconfident errors and improves selective generation performance on OOD benchmarks.
Self-Consistency Boosts Calibration for Math Reasoning (2024.findings-emnlp)

Copied to clipboard

Challenge: Existing solutions for math reasoning tasks use semantic parsing or AST decoding, but performance can degrade dramatically even with slight changes to the questions.
Approach: They propose three calibration methods based on self-consistency for math reasoning tasks.
Outcome: The proposed methods bridge model confidence and accuracy better than existing methods based on p(True) or logit.
Task Calibration: Calibrating Large Language Models on Inference Tasks (2025.findings-acl)

Copied to clipboard

Challenge: Large language models (LLMs) have shown impressive zero-shot performance on inference tasks, however, they may suffer from spurious correlations between input texts and output labels, which limits their ability to reason based purely on general language understanding.
Approach: They propose a zero-shot and inference-only calibration method inspired by mutual information which recovers LLM performance through task reformulation.
Outcome: The proposed calibration method improves on 13 benchmarks and prompt templates and can be integrated with other calibration methods.
Relevance Scores Calibration for Ranked List Truncation via TMP Adapter (2025.findings-acl)

Copied to clipboard

Challenge: ranked list truncation methods struggle with limited capacity, unstable training and inconsistency of selected thresholds.
Approach: They propose a new approach that incorporates the Threshold Margin Penalty as an additive loss function to calibrate ranking model relevance scores for ranked list truncation.
Outcome: The proposed method improves on retrieval datasets and offers theoretical and practical benefits.
DisCal: Distribution-Aware Calibration for Mathematical Reasoning Under Character-Level Noisy Inputs (2026.acl-long)

Copied to clipboard

Challenge: Existing methods for calibration of large reasoning models (LRMs) focus on clean inputs, leaving noise unexplored.
Approach: They propose a confidence calibration framework for character-level noisy inputs that extracts uncertainty signals from both the empirical answer distribution and the model’s predictive distribution and integrates them via a learned calibrator.
Outcome: Experiments on multiple mathematical reasoning benchmarks show that DisCal outperforms existing calibration methods under noisy inputs, reducing expected calibration error (ECE) by up to 39.21% and improving Area Under the Receiver Operating Characteristic Curve (AUROC) by 31.44%.
Calibrating Long-form Generations From Large Language Models (2024.findings-emnlp)

Copied to clipboard

Challenge: Conventional calibration methods treat answer correctness as binary and do not work for long-form generation where an answer can be partially correct.
Approach: They propose a framework where correctness of LLMs' responses and associated confidence levels are treated as distributions across a range of scores.
Outcome: The proposed framework treats the correctness of the LLMs’ responses and their associated confidence levels as distributions across a range of scores.
Fill In The Gaps: Model Calibration and Generalization with Synthetic Data (2024.emnlp-main)

Copied to clipboard

Challenge: Existing calibration methods negatively impact model accuracy due to the lack of diversity of validation data.
Approach: They propose a calibration method that incorporates synthetic data without compromising accuracy.
Outcome: The proposed method improves model accuracy on real data and reduces calibration error by 34% on four different tasks.
Similarity-Distance-Magnitude Activations (2026.findings-acl)

Copied to clipboard

Challenge: Existing methods for interpreting neural network-based language models (LMs) are limited to approximately conditional quantities.
Approach: They introduce a similarity-distance-magnitude activation function and an SDM estimator to control class- and prediction-conditional accuracy among selective classifications.
Outcome: The proposed estimator is more robust to covariate shifts and out-of-distribution inputs while remaining informative over in-difference data.
Beyond Generation: Leveraging LLM Creativity to Overcome Label Bias in Classification (2025.findings-acl)

Copied to clipboard

Challenge: Existing methods to mitigate label bias by leveraging in-domain data are often unavailable in real-world scenarios.
Approach: They propose a calibration method that generates synthetic in-domain data from a few in-context demonstrations and utilizes it for calibration.
Outcome: The proposed method reduces label bias by leveraging in-domain data from demonstrations.
KGE Calibrator: An Efficient Probability Calibration Method of Knowledge Graph Embedding Models for Trustworthy Link Prediction (2025.emnlp-main)

Copied to clipboard

Challenge: Existing methods for probability calibration of knowledge graph embedding models are ill-suited for KGEs.
Approach: They propose a method to calibrate knowledge graph embedding models for ranking-based link prediction using a Jump Selection Strategy and Multi-Binning Scaling to enhance reliability.
Outcome: Experiments show that the KGEC outperforms existing calibration methods in terms of effectiveness and efficiency.

What is GenGO?

GenGO is an NLP powered publication search system. It currenctly indexes 30k+ papers from ACL Anthology, and implements multi-aspect summarization, semantic search, and more!

Information

About
Limitations