# Formality is Favored: Unraveling the Learning Preferences of Large Language Models on Data with Conflicting Knowledge

arXiv: https://arxiv.org/abs/2410.04784 · HTML: https://mlsanity.com/paper/2410.04784/

## Key facts

- Cited by 8 papers in the map (data through 14 Oct 2026).
- Builds on 10 of its 16 references that are in the map.
- First posted on arXiv 7 Oct 2024.

## Builds on (most cited first, 10 of 10)

- [GPT-4 Technical Report](https://mlsanity.com/paper/2303.08774/) (2023)
- [Llama 2: Open Foundation and Fine-Tuned Chat Models](https://mlsanity.com/paper/2307.09288/) (2023)
- [Locating and Editing Factual Associations in GPT](https://mlsanity.com/paper/2202.05262/) (2022)
- [Large Language Models Encode Clinical Knowledge](https://mlsanity.com/paper/2212.13138/) (2022)
- [Pythia: A Suite for Analyzing Large Language Models Across Training and Scaling](https://mlsanity.com/paper/2304.01373/) (2023)
- [Transformer Feed-Forward Layers Are Key-Value Memories](https://mlsanity.com/paper/2012.14913/) (2020)
- [ChatGPT Outperforms Crowd-Workers for Text-Annotation Tasks](https://mlsanity.com/paper/2303.15056/) (2023)
- [Baichuan 2: Open Large-scale Language Models](https://mlsanity.com/paper/2309.10305/) (2023)
- [Harnessing the Power of LLMs in Practice: A Survey on ChatGPT and Beyond](https://mlsanity.com/paper/2304.13712/) (2023)
- [KGA: A General Machine Unlearning Framework Based on Knowledge Gap Alignment](https://mlsanity.com/paper/2305.06535/) (2023)

## Built on by (most cited first, 8 of 8)

- [Evaluating and Steering Modality Preferences in Multimodal Large Language Model](https://mlsanity.com/paper/2505.20977/) (2025)
- [Out of Style: RAG's Fragility to Linguistic Variation](https://mlsanity.com/paper/2504.08231/) (2025)
- [Where Knowledge Collides: A Mechanistic Study of Intra-Memory Knowledge Conflict in Language Models](https://mlsanity.com/paper/2601.09445/) (2026)
- [What Does Loss Optimization Actually Teach, If Anything? Knowledge Dynamics in Continual Pre-training of LLMs](https://mlsanity.com/paper/2601.03858/) (2026)
- [Truth as a Compression Artifact in Language Model Training](https://mlsanity.com/paper/2603.11749/) (2026)
- [Discrete Prompt Tuning via Recursive Utilization of Black-box Multimodal Large Language Model for Personalized Visual Emotion Recognition](https://mlsanity.com/paper/2509.04480/) (2025)
- [V-DyKnow: A Dynamic Benchmark for Time-Sensitive Knowledge in Vision Language Models](https://mlsanity.com/paper/2603.16581/) (2026)
- [Where's Waldo? Query-language Preference under Cross-lingual Knowledge Disparities](https://mlsanity.com/paper/2610.00606/) (2026)

## Abstract (arXiv)

Having been trained on massive pretraining data, large language models have shown excellent performance on many knowledge-intensive tasks. However, pretraining data tends to contain misleading and even conflicting information, and it is intriguing to understand how LLMs handle these noisy data during training. In this study, we systematically analyze LLMs' learning preferences for data with conflicting knowledge. We find that pretrained LLMs establish learning preferences similar to humans, i.e., preferences towards formal texts and texts with fewer spelling errors, resulting in faster learning and more favorable treatment of knowledge in data with such features when facing conflicts. This finding is generalizable across models and languages and is more evident in larger models. An in-depth analysis reveals that LLMs tend to trust data with features that signify consistency with the majority of data, and it is possible to instill new preferences and erase old ones by manipulating the degree of consistency with the majority data.