# Understanding and Mitigating Bias Inheritance in LLM-based Data Augmentation on Downstream Tasks

arXiv: https://arxiv.org/abs/2502.04419 · HTML: https://mlsanity.com/paper/2502.04419/

## Key facts

- Cited by 9 papers in the map (data through 14 Oct 2026).
- Placed in LLM synthetic data generation, one of 13 directions in Synthetic data and self-improvement for LLMs.
- Builds on 32 of its 62 references that are in the map.
- First posted on arXiv 6 Feb 2025.

## Placed in

- Synthetic data and self-improvement for LLMs / [LLM synthetic data generation](https://mlsanity.com/synth/llm-synthetic-data-generation/)

## Builds on (most cited first, 30 of 32)

- [LoRA: Low-Rank Adaptation of Large Language Models](https://mlsanity.com/paper/2106.09685/) (2021)
- [Man is to Computer Programmer as Woman is to Homemaker? Debiasing Word Embeddings](https://mlsanity.com/paper/1607.06520/) (2016)
- [Improving language models by retrieving from trillions of tokens](https://mlsanity.com/paper/2112.04426/) (2021)
- [Self-Play Fine-Tuning Converts Weak Language Models to Strong Language Models](https://mlsanity.com/paper/2401.01335/) (2024)
- [Bias in Bios: A Case Study of Semantic Representation Bias in a High-Stakes Setting](https://mlsanity.com/paper/1901.09451/) (2019)
- [Towards Understanding and Mitigating Social Biases in Language Models](https://mlsanity.com/paper/2106.13219/) (2021)
- [Physics of Language Models: Part 3.1, Knowledge Storage and Extraction](https://mlsanity.com/paper/2309.14316/) (2023)
- [Large Language Model as Attributed Training Data Generator: A Tale of Diversity and Bias](https://mlsanity.com/paper/2306.15895/) (2023)
- [A Pretrainer's Guide to Training Data: Measuring the Effects of Data Age, Domain Coverage, Quality, & Toxicity](https://mlsanity.com/paper/2305.13169/) (2023)
- [SWT-Bench: Testing and Validating Real-World Bug-Fixes with Code Agents](https://mlsanity.com/paper/2406.12952/) (2024)
- [A Note on Shumailov et al. (2024): `AI Models Collapse When Trained on Recursively Generated Data'](https://mlsanity.com/paper/2410.12954/) (2024)
- [To Repeat or Not To Repeat: Insights from Scaling LLM under Token-Crisis](https://mlsanity.com/paper/2305.13230/) (2023)
- [Contrasting Linguistic Patterns in Human and LLM-Generated News Text](https://mlsanity.com/paper/2308.09067/) (2023)
- ["It's a Fair Game", or Is It? Examining How Users Navigate Disclosure Risks and Benefits When Using LLM-Based Conversational Agents](https://mlsanity.com/paper/2309.11653/) (2023)
- [Toxic Language Detection in Social Media for Brazilian Portuguese: New Dataset and Multilingual Analysis](https://mlsanity.com/paper/2010.04543/) (2020)
- [Detection and Measurement of Syntactic Templates in Generated Text](https://mlsanity.com/paper/2407.00211/) (2024)
- ["You Gotta be a Doctor, Lin": An Investigation of Name-Based Bias of Large Language Models in Employment Recommendations](https://mlsanity.com/paper/2406.12232/) (2024)
- [Efficacy of Synthetic Data as a Benchmark](https://mlsanity.com/paper/2409.11968/) (2024)
- [Understanding and Mitigating the Label Noise in Pre-training on Downstream Tasks](https://mlsanity.com/paper/2309.17002/) (2023)
- [How to Synthesize Text Data without Model Collapse?](https://mlsanity.com/paper/2412.14689/) (2024)
- [A Robust Bias Mitigation Procedure Based on the Stereotype Content Model](https://mlsanity.com/paper/2210.14552/) (2022)
- [On Catastrophic Inheritance of Large Foundation Models](https://mlsanity.com/paper/2402.01909/) (2024)
- [OSACT4 Shared Task on Offensive Language Detection: Intensive Preprocessing-Based Approach](https://mlsanity.com/paper/2005.07297/) (2020)
- [Subtle Biases Need Subtler Measures: Dual Metrics for Evaluating Representative and Affinity Bias in Large Language Models](https://mlsanity.com/paper/2405.14555/) (2024)
- [Slight Corruption in Pre-training Data Makes Better Diffusion Models](https://mlsanity.com/paper/2405.20494/) (2024)
- [AI-UPV at IberLEF-2021 DETOXIS task: Toxicity Detection in Immigration-Related Web News Comments Using Transformers and Statistical Models](https://mlsanity.com/paper/2111.04530/) (2021)
- [Detect Camouflaged Spam Content via StoneSkipping: Graph and Text Joint Embedding for Chinese Character Variation Representation](https://mlsanity.com/paper/1908.11561/) (2019)
- [The Effects of Hallucinations in Synthetic Training Data for Relation Extraction](https://mlsanity.com/paper/2410.08393/) (2024)
- [Hey GPT, Can You be More Racist? Analysis from Crowdsourced Attempts to Elicit Biased Content from Generative AI](https://mlsanity.com/paper/2410.15467/) (2024)
- [Sina at FigNews 2024: Multilingual Datasets Annotated with Bias and Propaganda](https://mlsanity.com/paper/2407.09327/) (2024)

## Built on by (most cited first, 9 of 9)

- [Flattery in Motion: Benchmarking and Analyzing Sycophancy in Video-LLMs](https://mlsanity.com/paper/2506.07180/) (2025)
- [SynthRL: Scaling Visual Reasoning with Verifiable Data Synthesis](https://mlsanity.com/paper/2506.02096/) (2025)
- [Quantitative LLM Judges](https://mlsanity.com/paper/2506.02945/) (2025)
- [McBE: A Multi-task Chinese Bias Evaluation Benchmark for Large Language Models](https://mlsanity.com/paper/2507.02088/) (2025)
- [Safe for Whom? Rethinking How We Evaluate the Safety of LLMs for Real Users](https://mlsanity.com/paper/2512.10687/) (2025)
- [Cross-Language Bias Examination in Large Language Models](https://mlsanity.com/paper/2512.16029/) (2025)
- [Every Act Has Its Price: Compressed Moral Composition in Frontier LLMs](https://mlsanity.com/paper/2606.11232/) (2026)
- [RATE: An LLM-Powered Retrieval Augmented Generation Technology-Extraction Pipeline](https://mlsanity.com/paper/2507.21125/) (2025)
- ["Not in My Backyard": LLMs Uncover Online and Offline Social Biases Against Homelessness](https://mlsanity.com/paper/2508.13187/) (2025)

## Abstract (arXiv)

Generating synthetic datasets via large language models (LLMs) has emerged as a promising approach to improve LLM performance. However, LLMs inherently reflect biases in their training data, leading to a critical challenge: when models are trained on synthetic data, they may propagate and amplify the inherent biases that can significantly impact fairness and robustness on downstream tasks-a phenomenon we term bias inheritance. This work presents the first systematic investigation in understanding, analyzing, and mitigating bias inheritance. We fine-tune LLMs with a combined dataset of real and LLM-augmented data with varied bias ratio as the proportion of augmented data. Through systematic experiments across 10 classification and generation tasks, we analyze how 6 different types of biases manifest. Our results indicate that bias inheritance harms downstream task performance in bias directly-related classification and generation tasks. Then, our analysis identifies three key misalignment factors: misalignment of values, group data, and data distributions. Based on these insights, we propose three mitigation strategies: token-based, mask-based, and loss-based approaches, which can work differently on various tasks and bias, indicating the substantial challenges to mitigate bias inheritance. We hope this work can provide insights to the research of LLM data augmentation.