# The Unreasonable Effectiveness of Easy Training Data for Hard Tasks

arXiv: https://arxiv.org/abs/2401.06751 · HTML: https://mlsanity.com/paper/2401.06751/

## Key facts

- Cited by 40 papers in the map (data through 14 Oct 2026).
- Placed in Reasoning benchmarks and evaluation, one of 17 directions in LLM reasoning.
- Builds on 20 of its 48 references that are in the map.
- First posted on arXiv 12 Jan 2024.

## Placed in

- LLM reasoning / [Reasoning benchmarks and evaluation](https://mlsanity.com/lr/reasoning-benchmarks-and-evaluation/)

## Builds on (most cited first, 20 of 20)

- [Language Models are Few-Shot Learners](https://mlsanity.com/paper/2005.14165/) (2020)
- [PyTorch: An Imperative Style, High-Performance Deep Learning Library](https://mlsanity.com/paper/1912.01703/) (2019)
- [LLaMA: Open and Efficient Foundation Language Models](https://mlsanity.com/paper/2302.13971/) (2023)
- [Llama 2: Open Foundation and Fine-Tuned Chat Models](https://mlsanity.com/paper/2307.09288/) (2023)
- [Training Verifiers to Solve Math Word Problems](https://mlsanity.com/paper/2110.14168/) (2021)
- [Qwen Technical Report](https://mlsanity.com/paper/2309.16609/) (2023)
- [GPQA: A Graduate-Level Google-Proof Q&A Benchmark](https://mlsanity.com/paper/2311.12022/) (2023)
- [Mixtral of Experts](https://mlsanity.com/paper/2401.04088/) (2024)
- [Concrete Problems in AI Safety](https://mlsanity.com/paper/1606.06565/) (2016)
- [Probing Classifiers: Promises, Shortcomings, and Advances](https://mlsanity.com/paper/2102.12452/) (2021)
- [The Geometry of Truth: Emergent Linear Structure in Large Language Model Representations of True/False Datasets](https://mlsanity.com/paper/2310.06824/) (2023)
- [Faith and Fate: Limits of Transformers on Compositionality](https://mlsanity.com/paper/2305.18654/) (2023)
- [Weak-to-Strong Generalization: Eliciting Strong Capabilities With Weak Supervision](https://mlsanity.com/paper/2312.09390/) (2023)
- [Dataset Cartography: Mapping and Diagnosing Datasets with Training Dynamics](https://mlsanity.com/paper/2009.10795/) (2020)
- [Pervasive Label Errors in Test Sets Destabilize Machine Learning Benchmarks](https://mlsanity.com/paper/2103.14749/) (2021)
- [What Algorithms can Transformers Learn? A Study in Length Generalization](https://mlsanity.com/paper/2310.16028/) (2023)
- [Can You Learn an Algorithm? Generalizing from Easy to Hard Problems with Recurrent Networks](https://mlsanity.com/paper/2106.04537/) (2021)
- [End-to-end Algorithm Synthesis with Recurrent Networks: Logical Extrapolation Without Overthinking](https://mlsanity.com/paper/2202.05826/) (2022)
- [Rissanen Data Analysis: Examining Dataset Characteristics via Description Length](https://mlsanity.com/paper/2103.03872/) (2021)
- [Can Language Models Teach Weaker Agents? Teacher Explanations Improve Students via Personalization](https://mlsanity.com/paper/2306.09299/) (2023)

## Built on by (most cited first, 30 of 40)

- [Sparse Feature Circuits: Discovering and Editing Interpretable Causal Graphs in Language Models](https://mlsanity.com/paper/2403.19647/) (2024)
- [Easy-to-Hard Generalization: Scalable Alignment Beyond Human Supervision](https://mlsanity.com/paper/2403.09472/) (2024)
- [Can Language Models Learn to Skip Steps?](https://mlsanity.com/paper/2411.01855/) (2024)
- [Weak-to-Strong Reasoning](https://mlsanity.com/paper/2407.13647/) (2024)
- [Self-Improving Transformers Overcome Easy-to-Hard and Length Generalization Challenges](https://mlsanity.com/paper/2502.01612/) (2025)
- [The Delta Learning Hypothesis: Preference Tuning on Weak Data can Yield Strong Gains](https://mlsanity.com/paper/2507.06187/) (2025)
- [Easy2Hard-Bench: Standardized Difficulty Labels for Profiling LLM Performance and Generalization](https://mlsanity.com/paper/2409.18433/) (2024)
- [Your Weak LLM is Secretly a Strong Teacher for Alignment](https://mlsanity.com/paper/2409.08813/) (2024)
- [Do Language Models Mirror Human Confidence? Exploring Psychological Insights to Address Overconfidence in LLMs](https://mlsanity.com/paper/2506.00582/) (2025)
- [Provable Weak-to-Strong Generalization via Benign Overfitting](https://mlsanity.com/paper/2410.04638/) (2024)
- [Decoding Dark Matter: Specialized Sparse Autoencoders for Interpreting Rare Concepts in Foundation Models](https://mlsanity.com/paper/2411.00743/) (2024)
- [A transfer learning framework for weak-to-strong generalization](https://mlsanity.com/paper/2405.16236/) (2024)
- [The Road to Artificial SuperIntelligence: A Comprehensive Survey of Superalignment](https://mlsanity.com/paper/2412.16468/) (2024)
- [MathGAP: Out-of-Distribution Evaluation on Problems with Arbitrarily Complex Proofs](https://mlsanity.com/paper/2410.13502/) (2024)
- [Unsupervised Elicitation of Language Models](https://mlsanity.com/paper/2506.10139/) (2025)
- [Can Large Language Models Always Solve Easy Problems if They Can Solve Harder Ones?](https://mlsanity.com/paper/2406.12809/) (2024)
- [Varying Shades of Wrong: Aligning LLMs with Wrong Answers Only](https://mlsanity.com/paper/2410.11055/) (2024)
- [Improving Black-box Robustness with In-Context Rewriting](https://mlsanity.com/paper/2402.08225/) (2024)
- [Revisiting Generalization Across Difficulty Levels: It's Not So Easy](https://mlsanity.com/paper/2511.21692/) (2025)
- [Aletheia: What Makes RLVR For Code Verifiers Tick?](https://mlsanity.com/paper/2601.12186/) (2026)
- [EnsemW2S: Enhancing Weak-to-Strong Generalization with Large Language Model Ensembles](https://mlsanity.com/paper/2505.21959/) (2025)
- [P3: A Policy-Driven, Pace-Adaptive, and Diversity-Promoted Framework for data pruning in LLM Training](https://mlsanity.com/paper/2408.05541/) (2024)
- [How to Mitigate Overfitting in Weak-to-strong Generalization?](https://mlsanity.com/paper/2503.04249/) (2025)
- [Learning Composable Chains-of-Thought](https://mlsanity.com/paper/2505.22635/) (2025)
- [ReSURE: Regularizing Supervision Unreliability for Multi-turn Dialogue Fine-tuning](https://mlsanity.com/paper/2508.19996/) (2025)
- [Ranking Entities along Conceptual Space Dimensions with LLMs: An Analysis of Fine-Tuning Strategies](https://mlsanity.com/paper/2402.15337/) (2024)
- [Evaluating Fine-Tuning Efficiency of Human-Inspired Learning Strategies in Medical Question Answering](https://mlsanity.com/paper/2408.07888/) (2024)
- [Guiding Through Complexity: What Makes Good Supervision for Hard Math Reasoning Tasks?](https://mlsanity.com/paper/2410.20533/) (2024)
- [FLAMES: Improving LLM Math Reasoning via a Fine-Grained Analysis of the Data Synthesis Pipeline](https://mlsanity.com/paper/2508.16514/) (2025)
- [EnsemW2S: Enhancing Weak-to-Strong Generalization with Large Language Model Ensembles](https://mlsanity.com/paper/2410.04571/) (2024)

## Abstract (arXiv)

How can we train models to perform well on hard test data when hard training data is by definition difficult to label correctly? This question has been termed the scalable oversight problem and has drawn increasing attention as language models have continually improved. In this paper, we present the surprising conclusion that current pretrained language models often generalize relatively well from easy to hard data, even performing as well as oracle models finetuned on hard data. We demonstrate this kind of easy-to-hard generalization using simple finetuning methods like in-context learning, linear classifier heads, and QLoRA for seven different measures of datapoint hardness, including six empirically diverse human hardness measures (like grade level) and one model-based measure (loss-based). Furthermore, we show that even if one cares most about model performance on hard data, it can be better to collect easy data rather than hard data for finetuning, since hard data is generally noisier and costlier to collect. Our experiments use open models up to 70b in size and four publicly available question-answering datasets with questions ranging in difficulty from 3rd grade science questions to college level STEM questions and general-knowledge trivia. We conclude that easy-to-hard generalization in LMs is surprisingly strong for the tasks studied. Our code is available at: https://github.com/allenai/easy-to-hard-generalization