# Teaching People LLM's Errors and Getting it Right

arXiv: https://arxiv.org/abs/2512.21422 · HTML: https://mlsanity.com/paper/2512.21422/

## Key facts

- Cited by 1 papers in the map (data through 14 Oct 2026).
- Builds on 23 of its 80 references that are in the map.
- First posted on arXiv 24 Dec 2025.

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

- [Faster R-CNN: Towards Real-Time Object Detection with Region Proposal Networks](https://mlsanity.com/paper/1506.01497/) (2015)
- [Judging LLM-as-a-Judge with MT-Bench and Chatbot Arena](https://mlsanity.com/paper/2306.05685/) (2023)
- [Measuring Massive Multitask Language Understanding](https://mlsanity.com/paper/2009.03300/) (2020)
- [WildChat: 1M ChatGPT Interaction Logs in the Wild](https://mlsanity.com/paper/2405.01470/) (2024)
- [Does the Whole Exceed its Parts? The Effect of AI Explanations on Complementary Team Performance](https://mlsanity.com/paper/2006.14779/) (2020)
- [Effect of Confidence and Explanation on Accuracy and Trust Calibration in AI-Assisted Decision Making](https://mlsanity.com/paper/2001.02114/) (2020)
- [On Human Predictions with Explanations and Predictions of Machine Learning Models: A Case Study on Deception Detection](https://mlsanity.com/paper/1811.07901/) (2018)
- [Domino: Discovering Systematic Errors with Cross-Modal Embeddings](https://mlsanity.com/paper/2203.14960/) (2022)
- ["Why is 'Chicago' deceptive?" Towards Building Model-Driven Tutorials for Humans](https://mlsanity.com/paper/2001.05871/) (2020)
- [Human-AI Collaboration via Conditional Delegation: A Case Study of Content Moderation](https://mlsanity.com/paper/2204.11788/) (2022)
- [Goal Driven Discovery of Distributional Differences via Language Descriptions](https://mlsanity.com/paper/2302.14233/) (2023)
- [When does dough become a bagel? Analyzing the remaining mistakes on ImageNet](https://mlsanity.com/paper/2205.04596/) (2022)
- [Describing Differences between Text Distributions with Natural Language](https://mlsanity.com/paper/2201.12323/) (2022)
- [Why Does ChatGPT Fall Short in Providing Truthful Answers?](https://mlsanity.com/paper/2304.10513/) (2023)
- [Large Language Models Help Humans Verify Truthfulness -- Except When They Are Convincingly Wrong](https://mlsanity.com/paper/2310.12558/) (2023)
- [Improving Human-AI Collaboration With Descriptions of AI Behavior](https://mlsanity.com/paper/2301.06937/) (2023)
- [Effective Human-AI Teams via Learned Natural Language Rules and Onboarding](https://mlsanity.com/paper/2311.01007/) (2023)
- [SEAL : Interactive Tool for Systematic Error Analysis and Labeling](https://mlsanity.com/paper/2210.05839/) (2022)
- [Experimental Observations of the Topology of Convolutional Neural Network Activations](https://mlsanity.com/paper/2212.00222/) (2022)
- [DISCERN: Decoding Systematic Errors in Natural Language for Text Classifiers](https://mlsanity.com/paper/2410.22239/) (2024)
- [RealHarm: A Collection of Real-World Language Model Application Failures](https://mlsanity.com/paper/2504.10277/) (2025)
- [VISLIX: An XAI Framework for Validating Vision Models with Slice Discovery and Analysis](https://mlsanity.com/paper/2505.03132/) (2025)
- [Explainable Mapper: Charting LLM Embedding Spaces Using Perturbation-Based Explanation and Verification Agents](https://mlsanity.com/paper/2507.18607/) (2025)

## Built on by (most cited first, 1 of 1)

- [Supporting Calibrated Reliance in Human-AI Collaboration: Different Strategies for Different Tasks](https://mlsanity.com/paper/2604.03237/) (2026)

## Abstract (arXiv)

People use large language models (LLMs) when they should not. This is partly because they see LLMs compose poems and answer intricate questions, so they understandably, but incorrectly, assume LLMs won't stumble on basic tasks like simple arithmetic. Prior work has tried to address this by clustering instance embeddings into regions where an LLM is likely to fail and automatically describing patterns in these regions. The found failure patterns are taught to users to mitigate their overreliance. Yet, this approach has not fully succeeded. In this analysis paper, we aim to understand why. We first examine whether the negative result stems from the absence of failure patterns. We group instances in two datasets by their meta-labels and evaluate an LLM's predictions on these groups. We then define criteria to flag groups that are sizable and where the LLM is error-prone, and find meta-label groups that meet these criteria. Their meta-labels are the LLM's failure patterns that could be taught to users, so they do exist. We next test whether prompting and embedding-based approaches can surface these known failures. Without this, users cannot be taught about them to reduce their overreliance. We find mixed results across methods, which could explain the negative result. Finally, we revisit the final metric that measures teaching effectiveness. We propose to assess a user's ability to effectively use the given failure patterns to anticipate when an LLM is error-prone. A user study shows a positive effect from teaching with this metric, unlike the human-AI team accuracy. Our findings show that teaching failure patterns could be a viable approach to mitigating overreliance, but success depends on better automated failure-discovery methods and using metrics like ours.