# Data Curation Alone Can Stabilize In-context Learning

arXiv: https://arxiv.org/abs/2212.10378 · HTML: https://mlsanity.com/paper/2212.10378/

## Key facts

- Cited by 59 papers in the map (data through 14 Oct 2026).
- Builds on 38 of its 54 references that are in the map.
- First posted on arXiv 20 Dec 2022.

## Builds on (most cited first, 30 of 38)

- [BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding](https://mlsanity.com/paper/1810.04805/) (2018)
- [Language Models are Few-Shot Learners](https://mlsanity.com/paper/2005.14165/) (2020)
- [Sentence-BERT: Sentence Embeddings using Siamese BERT-Networks](https://mlsanity.com/paper/1908.10084/) (2019)
- [PaLM: Scaling Language Modeling with Pathways](https://mlsanity.com/paper/2204.02311/) (2022)
- [The Power of Scale for Parameter-Efficient Prompt Tuning](https://mlsanity.com/paper/2104.08691/) (2021)
- [OPT: Open Pre-trained Transformer Language Models](https://mlsanity.com/paper/2205.01068/) (2022)
- [Scaling Instruction-Finetuned Language Models](https://mlsanity.com/paper/2210.11416/) (2022)
- [UMAP: Uniform Manifold Approximation and Projection for Dimension Reduction](https://mlsanity.com/paper/1802.03426/) (2018)
- [Emergent Abilities of Large Language Models](https://mlsanity.com/paper/2206.07682/) (2022)
- [Character-level Convolutional Networks for Text Classification](https://mlsanity.com/paper/1509.01626/) (2015)
- [BoolQ: Exploring the Surprising Difficulty of Natural Yes/No Questions](https://mlsanity.com/paper/1905.10044/) (2019)
- [Understanding Black-box Predictions via Influence Functions](https://mlsanity.com/paper/1703.04730/) (2017)
- [Multitask Prompted Training Enables Zero-Shot Task Generalization](https://mlsanity.com/paper/2110.08207/) (2021)
- [Rethinking the Role of Demonstrations: What Makes In-Context Learning Work?](https://mlsanity.com/paper/2202.12837/) (2022)
- [Calibrate Before Use: Improving Few-Shot Performance of Language Models](https://mlsanity.com/paper/2102.09690/) (2021)
- [Making Pre-trained Language Models Better Few-shot Learners](https://mlsanity.com/paper/2012.15723/) (2020)
- [LaMDA: Language Models for Dialog Applications](https://mlsanity.com/paper/2201.08239/) (2022)
- [Scaling Language Models: Methods, Analysis & Insights from Training Gopher](https://mlsanity.com/paper/2112.11446/) (2021)
- [What Makes Good In-Context Examples for GPT-3?](https://mlsanity.com/paper/2101.06804/) (2021)
- [Fantastically Ordered Prompts and Where to Find Them: Overcoming Few-Shot Prompt Order Sensitivity](https://mlsanity.com/paper/2104.08786/) (2021)
- [Exploiting Cloze Questions for Few Shot Text Classification and Natural Language Inference](https://mlsanity.com/paper/2001.07676/) (2020)
- [Learning To Retrieve Prompts for In-Context Learning](https://mlsanity.com/paper/2112.08633/) (2021)
- [Data Shapley: Equitable Valuation of Data for Machine Learning](https://mlsanity.com/paper/1904.02868/) (2019)
- [Estimating Training Data Influence by Tracing Gradient Descent](https://mlsanity.com/paper/2002.08484/) (2020)
- [Using DeepSpeed and Megatron to Train Megatron-Turing NLG 530B, A Large-Scale Generative Language Model](https://mlsanity.com/paper/2201.11990/) (2022)
- [True Few-Shot Learning with Language Models](https://mlsanity.com/paper/2105.11447/) (2021)
- [PromptSource: An Integrated Development Environment and Repository for Natural Language Prompts](https://mlsanity.com/paper/2202.01279/) (2022)
- [Selective Annotation Makes Language Models Better Few-Shot Learners](https://mlsanity.com/paper/2209.01975/) (2022)
- [Demystifying Prompts in Language Models via Perplexity Estimation](https://mlsanity.com/paper/2212.04037/) (2022)
- [Active Example Selection for In-Context Learning](https://mlsanity.com/paper/2211.04486/) (2022)

## Built on by (most cited first, 30 of 59)

- [Advances and Challenges in Foundation Agents: From Brain-Inspired Intelligence to Evolutionary, Collaborative, and Safe Systems](https://mlsanity.com/paper/2504.01990/) (2025)
- [Finding Support Examples for In-Context Learning](https://mlsanity.com/paper/2302.13539/) (2023)
- [ContextCite: Attributing Model Generation to Context](https://mlsanity.com/paper/2409.00729/) (2024)
- [Mind Your Format: Towards Consistent Evaluation of In-Context Learning Improvements](https://mlsanity.com/paper/2401.06766/) (2024)
- [Breaking Agents: Compromising Autonomous LLM Agents Through Malfunction Amplification](https://mlsanity.com/paper/2407.20859/) (2024)
- [Transformers Learn to Achieve Second-Order Convergence Rates for In-Context Linear Regression](https://mlsanity.com/paper/2310.17086/) (2023)
- [IsoBench: Benchmarking Multimodal Foundation Models on Isomorphic Representations](https://mlsanity.com/paper/2404.01266/) (2024)
- [Prompt Optimization with EASE? Efficient Ordering-aware Automated Selection of Exemplars](https://mlsanity.com/paper/2405.16122/) (2024)
- [How are Prompts Different in Terms of Sensitivity?](https://mlsanity.com/paper/2311.07230/) (2023)
- [Get more for less: Principled Data Selection for Warming Up Fine-Tuning in LLMs](https://mlsanity.com/paper/2405.02774/) (2024)
- [Smoothie: Label Free Language Model Routing](https://mlsanity.com/paper/2412.04692/) (2024)
- [A Survey of Table Reasoning with Large Language Models](https://mlsanity.com/paper/2402.08259/) (2024)
- [IDEAL: Influence-Driven Selective Annotations Empower In-Context Learners in Large Language Models](https://mlsanity.com/paper/2310.10873/) (2023)
- [Beyond Performance: Quantifying and Mitigating Label Bias in LLMs](https://mlsanity.com/paper/2405.02743/) (2024)
- [RetICL: Sequential Retrieval of In-Context Examples with Reinforcement Learning](https://mlsanity.com/paper/2305.14502/) (2023)
- [LLM Task Interference: An Initial Study on the Impact of Task-Switch in Conversational History](https://mlsanity.com/paper/2402.18216/) (2024)
- [Stress-Testing Long-Context Language Models with Lifelong ICL and Task Haystack](https://mlsanity.com/paper/2407.16695/) (2024)
- [Comparing Specialised Small and General Large Language Models on Text Classification: 100 Labelled Samples to Achieve Break-Even Performance](https://mlsanity.com/paper/2402.12819/) (2024)
- [Effects of diversity incentives on sample diversity and downstream model performance in LLM-based text augmentation](https://mlsanity.com/paper/2401.06643/) (2024)
- [SynthDST: Synthetic Data is All You Need for Few-Shot Dialog State Tracking](https://mlsanity.com/paper/2402.02285/) (2024)
- [Revisiting In-context Learning Inference Circuit in Large Language Models](https://mlsanity.com/paper/2410.04468/) (2024)
- [Reconstruct Your Previous Conversations! Comprehensively Investigating Privacy Leakage Risks in Conversations with GPT Models](https://mlsanity.com/paper/2402.02987/) (2024)
- [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)
- [Improving Probability-based Prompt Selection Through Unified Evaluation and Analysis](https://mlsanity.com/paper/2305.14877/) (2023)
- [In-Context Example Ordering Guided by Label Distributions](https://mlsanity.com/paper/2402.11447/) (2024)
- [Addressing Order Sensitivity of In-Context Demonstration Examples in Causal Language Models](https://mlsanity.com/paper/2402.15637/) (2024)
- [In-Context Learning Demonstration Selection via Influence Analysis](https://mlsanity.com/paper/2402.11750/) (2024)
- [Strategic Demonstration Selection for Improved Fairness in LLM In-Context Learning](https://mlsanity.com/paper/2408.09757/) (2024)
- [FastGAS: Fast Graph-based Annotation Selection for In-Context Learning](https://mlsanity.com/paper/2406.03730/) (2024)
- [LLMs vs Established Text Augmentation Techniques for Classification: When do the Benefits Outweight the Costs?](https://mlsanity.com/paper/2408.16502/) (2024)

## Abstract (arXiv)

In-context learning (ICL) enables large language models (LLMs) to perform new tasks by prompting them with a sequence of training examples. However, it is known that ICL is very sensitive to the choice of training examples: randomly sampling examples from a training set leads to high variance in performance. In this paper, we show that carefully curating a subset of training data greatly stabilizes ICL performance without any other changes to the ICL algorithm (e.g., prompt retrieval or calibration). We introduce two methods to choose training subsets -- both score training examples individually, then select the highest-scoring ones. CondAcc scores a training example by its average dev-set ICL accuracy when combined with random training examples, while Datamodels learns linear regressors that estimate how the presence of each training example influences LLM outputs. Across five tasks and two LLMs, sampling from stable subsets selected by CondAcc and Datamodels improves average accuracy over sampling from the entire training set by 7.7% and 6.3%, respectively. Surprisingly, the stable subset examples are not especially diverse in content or low in perplexity, in contrast with other work suggesting that diversity and perplexity are important when prompting LLMs.