# Are Larger Pretrained Language Models Uniformly Better? Comparing Performance at the Instance Level

arXiv: https://arxiv.org/abs/2105.06020 · HTML: https://mlsanity.com/paper/2105.06020/

## Key facts

- Cited by 34 papers in the map (data through 14 Oct 2026).
- Builds on 25 of its 32 references that are in the map.
- First posted on arXiv 13 May 2021.

## Builds on (most cited first, 25 of 25)

- [BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding](https://mlsanity.com/paper/1810.04805/) (2018)
- [Exploring the Limits of Transfer Learning with a Unified Text-to-Text Transformer](https://mlsanity.com/paper/1910.10683/) (2019)
- [Scaling Laws for Neural Language Models](https://mlsanity.com/paper/2001.08361/) (2020)
- [BART: Denoising Sequence-to-Sequence Pre-training for Natural Language Generation, Translation, and Comprehension](https://mlsanity.com/paper/1910.13461/) (2019)
- [GLUE: A Multi-Task Benchmark and Analysis Platform for Natural Language Understanding](https://mlsanity.com/paper/1804.07461/) (2018)
- [SQuAD: 100,000+ Questions for Machine Comprehension of Text](https://mlsanity.com/paper/1606.05250/) (2016)
- [Switch Transformers: Scaling to Trillion Parameter Models with Simple and Efficient Sparsity](https://mlsanity.com/paper/2101.03961/) (2021)
- [Deep contextualized word representations](https://mlsanity.com/paper/1802.05365/) (2018)
- [A large annotated corpus for learning natural language inference](https://mlsanity.com/paper/1508.05326/) (2015)
- [Beyond Accuracy: Behavioral Testing of NLP models with CheckList](https://mlsanity.com/paper/2005.04118/) (2020)
- [Right for the Wrong Reasons: Diagnosing Syntactic Heuristics in Natural Language Inference](https://mlsanity.com/paper/1902.01007/) (2019)
- [Designing and Interpreting Probes with Control Tasks](https://mlsanity.com/paper/1909.03368/) (2019)
- [Scaling Laws for Autoregressive Generative Modeling](https://mlsanity.com/paper/2010.14701/) (2020)
- [Underspecification Presents Challenges for Credibility in Modern Machine Learning](https://mlsanity.com/paper/2011.03395/) (2020)
- [Fine-Tuning Pretrained Language Models: Weight Initializations, Data Orders, and Early Stopping](https://mlsanity.com/paper/2002.06305/) (2020)
- [Calibration of Pre-trained Transformers](https://mlsanity.com/paper/2003.07892/) (2020)
- [An Investigation of Why Overparameterization Exacerbates Spurious Correlations](https://mlsanity.com/paper/2005.04345/) (2020)
- [Pretrained Transformers Improve Out-of-Distribution Robustness](https://mlsanity.com/paper/2004.06100/) (2020)
- [Stress Test Evaluation for Natural Language Inference](https://mlsanity.com/paper/1806.00692/) (2018)
- [What Can We Learn from Collective Human Opinions on Natural Language Inference Data?](https://mlsanity.com/paper/2010.03532/) (2020)
- [Intermediate-Task Transfer Learning with Pretrained Models for Natural Language Understanding: When and Why Does It Work?](https://mlsanity.com/paper/2005.00628/) (2020)
- [BERTs of a feather do not generalize together: Large variability in generalization across models with similar test set performance](https://mlsanity.com/paper/1911.02969/) (2019)
- [A Continuously Growing Dataset of Sentential Paraphrases](https://mlsanity.com/paper/1708.00391/) (2017)
- [Train Large, Then Compress: Rethinking Model Size for Efficient Training and Inference of Transformers](https://mlsanity.com/paper/2002.11794/) (2020)
- [Predicting Declension Class from Form and Meaning](https://mlsanity.com/paper/2005.00626/) (2020)

## Built on by (most cited first, 30 of 34)

- [What Does it Mean for a Language Model to Preserve Privacy?](https://mlsanity.com/paper/2202.05520/) (2022)
- [Datamodels: Predicting Predictions from Training Data](https://mlsanity.com/paper/2202.00622/) (2022)
- [The MultiBERTs: BERT Reproductions for Robustness Analysis](https://mlsanity.com/paper/2106.16163/) (2021)
- [Anchor Points: Benchmarking Models with Much Fewer Examples](https://mlsanity.com/paper/2309.08638/) (2023)
- [TopViewRS: Vision-Language Models as Top-View Spatial Reasoners](https://mlsanity.com/paper/2406.02537/) (2024)
- [Model Cascading: Towards Jointly Improving Efficiency and Accuracy of NLP Systems](https://mlsanity.com/paper/2210.05528/) (2022)
- [ModelDiff: A Framework for Comparing Learning Algorithms](https://mlsanity.com/paper/2211.12491/) (2022)
- [Grounding Representation Similarity with Statistical Testing](https://mlsanity.com/paper/2108.01661/) (2021)
- [It Takes Two to Tango: Navigating Conceptualizations of NLP Tasks and Measurements of Performance](https://mlsanity.com/paper/2305.09022/) (2023)
- [Scaling Laws Under the Microscope: Predicting Transformer Performance from Small Scale Experiments](https://mlsanity.com/paper/2202.06387/) (2022)
- [Deconstructing Distributions: A Pointwise Framework of Learning](https://mlsanity.com/paper/2202.09931/) (2022)
- [Evaluating Distributional Distortion in Neural Language Modeling](https://mlsanity.com/paper/2203.12788/) (2022)
- [ILDAE: Instance-Level Difficulty Analysis of Evaluation Data](https://mlsanity.com/paper/2203.03073/) (2022)
- [GULP: a prediction-based metric between representations](https://mlsanity.com/paper/2210.06545/) (2022)
- ["John is 50 years old, can his son be 65?" Evaluating NLP Models' Understanding of Feasibility](https://mlsanity.com/paper/2210.07471/) (2022)
- [MenatQA: A New Dataset for Testing the Temporal Comprehension and Reasoning Abilities of Large Language Models](https://mlsanity.com/paper/2310.05157/) (2023)
- [A Survey on Stability of Learning with Limited Labelled Data and its Sensitivity to the Effects of Randomness](https://mlsanity.com/paper/2312.01082/) (2023)
- [Understanding the Mechanisms of Fast Hyperparameter Transfer](https://mlsanity.com/paper/2512.22768/) (2025)
- [Predicting on the Edge: Identifying Where a Larger Model Does Better](https://mlsanity.com/paper/2202.07652/) (2022)
- [Analyzing Semantic Faithfulness of Language Models via Input Intervention on Question Answering](https://mlsanity.com/paper/2212.10696/) (2022)
- [All Roads Lead to Rome? Exploring the Invariance of Transformers' Representations](https://mlsanity.com/paper/2305.14555/) (2023)
- [How Emotionally Stable is ALBERT? Testing Robustness with Stochastic Weight Averaging on a Sentiment Analysis Task](https://mlsanity.com/paper/2111.09612/) (2021)
- [The Effect of Model Size on Worst-Group Generalization](https://mlsanity.com/paper/2112.04094/) (2021)
- [Beyond Mahalanobis-Based Scores for Textual OOD Detection](https://mlsanity.com/paper/2211.13527/) (2022)
- [Learning to Rank for Active Learning via Multi-Task Bilevel Optimization](https://mlsanity.com/paper/2310.17044/) (2023)
- [On Affine Homotopy between Language Encoders](https://mlsanity.com/paper/2406.02329/) (2024)
- [Building Legal Datasets](https://mlsanity.com/paper/2111.02034/) (2021)
- [Synergy and Diversity in CLIP: Enhancing Performance Through Adaptive Backbone Ensembling](https://mlsanity.com/paper/2405.17139/) (2024)
- [DICE: Structured Reasoning in LLMs through SLM-Guided Chain-of-Thought Correction](https://mlsanity.com/paper/2510.09211/) (2025)
- [Cross-Model Comparative Loss for Enhancing Neuronal Utility in Language Understanding](https://mlsanity.com/paper/2301.03765/) (2023)

## Abstract (arXiv)

Larger language models have higher accuracy on average, but are they better on every single instance (datapoint)? Some work suggests larger models have higher out-of-distribution robustness, while other work suggests they have lower accuracy on rare subgroups. To understand these differences, we investigate these models at the level of individual instances. However, one major challenge is that individual predictions are highly sensitive to noise in the randomness in training. We develop statistically rigorous methods to address this, and after accounting for pretraining and finetuning noise, we find that our BERT-Large is worse than BERT-Mini on at least 1-4% of instances across MNLI, SST-2, and QQP, compared to the overall accuracy improvement of 2-10%. We also find that finetuning noise increases with model size and that instance-level accuracy has momentum: improvement from BERT-Mini to BERT-Medium correlates with improvement from BERT-Medium to BERT-Large. Our findings suggest that instance-level predictions provide a rich source of information; we therefore, recommend that researchers supplement model weights with model predictions.