Removing personally identifiable information (PII) from texts is necessary to comply with various data protection regulations and to enable data sharing without compromising privacy. However, recent works show that documents sanitized by PII-removal techniques are vulnerable to reconstruction attacks. Yet, we suspect that the reported success of these attacks is largely overestimated. We critically analyze the evaluation of existing attacks and find that data leakage and data contamination are not properly mitigated, leaving the question whether or not PII removal techniques truly protect privacy in real-world scenarios unaddressed. We investigate possible data sources and attack setups that avoid data leakage and conclude that only truly private data can allow us to objectively evaluate vulnerabilities in PII removal techniques. However, access to private data is heavily restricted—and for good reasons—which also means that the public research community cannot address this problem in a transparent, reproducible, and trustworthy manner.

How capable are diffusion models of generating synthetics texts? Recent research shows their strengths, with performance reaching that of auto-regressive LLMs. But are they also good in generating synthetic data if the training was under differential privacy? Here the evidence is missing, yet the promises from private image generation look strong. In this paper we address this open question by extensive experiments. At the same time, we critically assess (and reimplement) previous works on synthetic private text generation with LLMs and reveal some unmet assumptions that might have led to violating the differential privacy guarantees. Our results partly contradict previous non-private findings and show that fully open-source LLMs outperform diffusion models in the privacy regime. Our complete source codes, datasets, and experimental setup is publicly available to foster future research.

Handing in a paper or exercise and merely receiving “bad” or “incorrect” as feedback is not very helpful when the goal is to improve. Unfortunately, this is currently the kind of feedback given by Automatic Short Answer Grading (ASAG) systems. One of the reasons for this is a lack of content-focused elaborated feedback datasets. To encourage research on explainable and understandable feedback systems, we present the Short Answer Feedback dataset (SAF). Similar to other ASAG datasets, SAF contains learner responses and reference answers to German and English questions. However, instead of only assigning a label or score to the learners’ answers, SAF also contains elaborated feedback explaining the given score. Thus, SAF enables supervised training of models that grade answers and explain where and why mistakes were made. This paper discusses the need for enhanced feedback models in real-world pedagogical scenarios, describes the dataset annotation process, gives a comprehensive analysis of SAF, and provides T5-based baselines for future comparison.