The paper investigates how biases present in in-context examples used by Large Language Models (LLMs) for synthetic tabular data generation can propagate to the generated data, causing global statistical distortions even when the biases are mild. It also explores an adversarial scenario where a malicious actor can inject bias through a subset of examples, thereby compromising the fairness of downstream classifiers for protected subgroups. The authors evaluate prompt-side mitigation strategies, such as preprocessing examples and instruction-based defenses, noting that while these can reduce disparity, LLMs remain highly sensitive to adversarial prompts, revealing a significant vulnerability in data generation pipelines for sensitive domains.
arXiv:2509. 09960v2 Announce Type: replace-cross Abstract: Synthetic tabular data generation is increasingly essential in machine learning, supporting downstream applications when real-world, high-quality tabular data is insufficient.
arXiv:2610.16990v1 Announce Type: cross Abstract: Tabular data generation aims to synthesize realistic samples from limited observations while faithfully capturing the underlying data distribution. T...
The paper examines how synthetic data can be used to covertly inject targeted social biases into aligned large language models. By having a misaligned teacher model generate filtered synthetic datasets in domains like creative writing and code generation, the authors fine‑tune aligned student models and demonstrate that the benign‑looking data can transmit biases while preserving general task performance. The study highlights a security risk in synthetic data‑driven LLM training pipelines and proposes log‑linearity‑based scoring as a potential safeguard.
The paper introduces FAB, an attack that uses meta‑learning to embed dormant adversarial behaviors into large language models (LLMs). These behaviors remain inactive until the model is finetuned by downstream users, at which point the model can exhibit unwanted actions such as unsolicited advertising, jailbreakability, or over‑refusal. FAB is shown to be effective across multiple LLMs and resilient to various finetuning settings.
arXiv:2607. 13541v1 Announce Type: cross Abstract: To overcome data scarcity and privacy constraints in data collection, it has become standard practice across academia and industry to augment real training data with text-to-image (T2I)-generated synthetic data, a paradigm we term Real-Synthetic Mix-Training (RSMT).
arXiv:2606. 18307v1 Announce Type: cross Abstract: Optimizing the training data distribution for Supervised Fine-Tuning (SFT) dictates the capability of Large Language Models (LLMs).
INCONVENIENT is a new benchmark that evaluates how machine‑learning models, especially large language models, generalize when trained on convenience samples that bias feature–target relationships. Unlike prior distribution‑shift tests that rely on natural domain splits, INCONVENIENT explicitly perturbs these relationships while keeping the deployment population unchanged, allowing controlled study of shift severity and direction. The benchmark covers 200 tabular classification and regression datasets and produces about 15,500 shifted training distributions, revealing significant performance degradation.
arXiv:2606. 02947v1 Announce Type: new Abstract: Supervised fine-tuning is the predominant approach for adapting autoregressive vision-language models to downstream tasks.
arXiv:2608. 04286v1 Announce Type: cross Abstract: Large language models (LLMs) are often used in conjunction with external knowledge sources to improve their factual accuracy and decrease hallucinations, through methods such as Retrieval-Augmented Generation (RAG).
TextCloak is a reinforcement‑learning driven framework that generates unlearnable examples (UEs) to protect textual data from unauthorized fine‑tuning of large language models (LLMs). It transforms clean text into perturbations that preserve meaning and naturalness while degrading the utility of models trained on them. Experiments on six datasets and nine LLMs show that TextCloak consistently impairs unauthorized fine‑tuning and remains robust across different architectures, training settings, and adaptive attacks.
arXiv:2603. 19423v2 Announce Type: replace-cross Abstract: Large language model (LLM) agents increasingly rely on external tools (file operations, API calls, database transactions) to autonomously complete complex multi-step tasks.
The paper introduces PropMe, a framework that evaluates both the capability of large language models (LLMs) to reproduce training data via prefix attacks and their natural propensity to do so without adversarial prompting. Using SimpleTrace to compare every generation against the full indexed training corpus, the authors test models such as Comma, DFM Decoder, and OLMo3 32B across datasets like Common Pile, Dynaword, and Dolma3. Results show that natural sentence prompts produce less training text than prefix attacks, but unprompted or single-word prompts can reproduce as much or more, often involving duplicated licenses, code, and templates; continual pre‑training only modestly reduces verbatim reproduction.