# CONTESTS: a Framework for Consistency Testing of Span Probabilities in Language Models

arXiv: https://arxiv.org/abs/2409.19984 · HTML: https://mlsanity.com/paper/2409.19984/

## Key facts

- Cited by 3 papers in the map (data through 14 Oct 2026).
- Builds on 23 of its 41 references that are in the map.
- First posted on arXiv 30 Sep 2024.

## Builds on (most cited first, 23 of 23)

- [Language Models are Few-Shot Learners](https://mlsanity.com/paper/2005.14165/) (2020)
- [Training language models to follow instructions with human feedback](https://mlsanity.com/paper/2203.02155/) (2022)
- [Exploring the Limits of Transfer Learning with a Unified Text-to-Text Transformer](https://mlsanity.com/paper/1910.10683/) (2019)
- [RoBERTa: A Robustly Optimized BERT Pretraining Approach](https://mlsanity.com/paper/1907.11692/) (2019)
- [On Calibration of Modern Neural Networks](https://mlsanity.com/paper/1706.04599/) (2017)
- [Unsupervised Cross-lingual Representation Learning at Scale](https://mlsanity.com/paper/1911.02116/) (2019)
- [Finetuned Language Models Are Zero-Shot Learners](https://mlsanity.com/paper/2109.01652/) (2021)
- [The Curious Case of Neural Text Degeneration](https://mlsanity.com/paper/1904.09751/) (2019)
- [Measuring and Improving Consistency in Pretrained Language Models](https://mlsanity.com/paper/2102.01017/) (2021)
- [Calibration of Pre-trained Transformers](https://mlsanity.com/paper/2003.07892/) (2020)
- [Theoretical Limitations of Self-Attention in Neural Sequence Models](https://mlsanity.com/paper/1906.06755/) (2019)
- [Efficient Training of Language Models to Fill in the Middle](https://mlsanity.com/paper/2207.14255/) (2022)
- [BERT has a Mouth, and It Must Speak: BERT as a Markov Random Field Language Model](https://mlsanity.com/paper/1902.04094/) (2019)
- [Breaking the Softmax Bottleneck: A High-Rank RNN Language Model](https://mlsanity.com/paper/1711.03953/) (2017)
- [Calibrating Sequence likelihood Improves Conditional Language Generation](https://mlsanity.com/paper/2210.00045/) (2022)
- [Trading Off Diversity and Quality in Natural Language Generation](https://mlsanity.com/paper/2004.10450/) (2020)
- [A Logic-Driven Framework for Consistency of Neural Models](https://mlsanity.com/paper/1909.00126/) (2019)
- [A Close Look into the Calibration of Pre-trained Language Models](https://mlsanity.com/paper/2211.00151/) (2022)
- [A Measure-Theoretic Characterization of Tight Language Models](https://mlsanity.com/paper/2212.10502/) (2022)
- [Conformal Nucleus Sampling](https://mlsanity.com/paper/2305.02633/) (2023)
- [Calibration, Entropy Rates, and Memory in Language Models](https://mlsanity.com/paper/1906.05664/) (2019)
- [Preserving Pre-trained Features Helps Calibrate Fine-tuned Language Models](https://mlsanity.com/paper/2305.19249/) (2023)
- [On the Discrepancy between Density Estimation and Sequence Generation](https://mlsanity.com/paper/2002.07233/) (2020)

## Built on by (most cited first, 3 of 3)

- [Probability Contracts: Accuracy, Coherence, and Decisions Across LLM Interfaces](https://mlsanity.com/paper/2609.37470/) (2026)
- [Express Your Doubts -- Probabilistic World Modeling Should not be Based on Token logprobs](https://mlsanity.com/paper/2505.02072/) (2025)
- [Path-Dependent Denoising: A Non-Conservative Field Perspective on Order Collapse in Diffusion Language Models](https://mlsanity.com/paper/2605.09303/) (2026)

## Abstract (arXiv)

Although language model scores are often treated as probabilities, their reliability as probability estimators has mainly been studied through calibration, overlooking other aspects. In particular, it is unclear whether language models produce the same value for different ways of assigning joint probabilities to word spans. Our work introduces a novel framework, ConTestS (Consistency Testing over Spans), involving statistical tests to assess score consistency across interchangeable completion and conditioning orders. We conduct experiments on post-release real and synthetic data to eliminate training effects. Our findings reveal that both Masked Language Models (MLMs) and autoregressive models exhibit inconsistent predictions, with autoregressive models showing larger discrepancies. Larger MLMs tend to produce more consistent predictions, while autoregressive models show the opposite trend. Moreover, for both model types, prediction entropies offer insights into the true word span likelihood and therefore can aid in selecting optimal decoding strategies. The inconsistencies revealed by our analysis, as well their connection to prediction entropies and differences between model types, can serve as useful guides for future research on addressing these limitations.