Quantized large language models (LLMs) have gained increasing attention and significance for enabling deployment in resource-constrained environments. However, emerging studies on a few calibration dataset-free quantization methods suggest that quantization may compromise the safety capabilities of LLMs, underscoring the urgent need for systematic safety evaluations and effective mitigation strategies. In this paper, we present comprehensive safety evaluations across various mainstream quantization techniques and diverse calibration datasets, utilizing widely accepted safety benchmarks. To address the identified safety vulnerabilities, we propose a quantization-aware safety patching framework, Q-resafe, to efficiently restore the safety capabilities of quantized LLMs while minimizing any adverse impact on utility. Extensive experiment results demonstrate that Q-resafe successfully re-aligns the safety of quantized LLMs with their pre-quantization counterparts, even under challenging evaluation scenarios. Project page: https://github.com/Thecommonirin/Qresafe.

Federated learning (FL) is predicated on the provision of high-quality data by multiple clients, which is then used to train global models. A plethora of incentive mechanism studies have been conducted with the objective of promoting the provision of high-quality data by clients. These studies have focused on the distribution of benefits to clients. However, the incentives of federated learning are transactional in nature, and the issue of the atomicity of transactions has not been addressed. Furthermore, the data quality of individual clients participating in training varies, and they may participate negatively in training out of privacy leakage concerns.Consequently, we propose an inaugural atomistic incentive scheme with privacy preservation in the FL setting: πFL (privacy, atomic, incentive). This scheme establishes a more dependable training environment based on Shapley valuation, secure multi-party computation, and smart contracts. Consequently, it ensures that each client's contribution can be accurately measured and appropriately rewarded, improves the accuracy and efficiency of model training, and enhances the sustainability and reliability of the FL system. The efficacy of this mechanism has been demonstrated through comprehensive experimental analysis. It is evident that this mechanism not only protects the privacy of trainers and provides atomic training rewards but also improves the model performance of FL, with an accuracy improvement of at least 8%.

With the growing popularity of LLMs among the general public users, privacy-preserving and adversarial robustness have become two pressing demands for LLM-based services, which have largely been pursued separately but rarely jointly. In this paper, to the best of our knowledge, we are among the first attempts towards robust and private LLM inference by tightly integrating two disconnected fields: private inference and prompt ensembling. The former protects users’ privacy by encrypting inference data transmitted and processed by LLMs, while the latter enhances adversarial robustness by yielding an aggregated output from multiple prompted LLM responses. Although widely recognized as effective individually, private inference for prompt ensembling together entails new challenges that render the naive combination of existing techniques inefficient. To overcome the hurdles, we propose SecPE, which designs efficient fully homomorphic encryption (FHE) counterparts for the core algorithmic building blocks of prompt ensembling. We conduct extensive experiments on 8 tasks to evaluate the accuracy, robustness, and efficiency of SecPE. The results show that SecPE maintains high clean accuracy and offers better robustness at the expense of merely 2.5% efficiency overhead compared to baseline private inference methods, indicating a satisfactory “accuracy-robustness-efficiency” tradeoff. For the efficiency of the encrypted Argmax operation that incurs major slowdown for prompt ensembling, SecPE is 35.4 times faster than the state-of-the-art peers, which can be of independent interest beyond this work.

Due to its powerful representational capabilities, Transformers have gradually become the mainstream model in the field of machine vision. However, the vast and complex parameters of Transformers impede researchers from gaining a deep understanding of their internal mechanisms, especially error mechanisms. Existing methods for interpreting Transformers mainly focus on understanding them from the perspectives of the importance of input tokens or internal modules, as well as the formation and meaning of features. In contrast, inspired by research on information integration mechanisms and conjunctive errors in the biological visual system, this paper conducts an in-depth exploration of the internal error mechanisms of Transformers. We first propose an information integration hypothesis for Transformers in the machine vision domain and provide substantial experimental evidence to support this hypothesis. This includes the dynamic integration of information among tokens and the static integration of information within tokens in Transformers, as well as the presence of conjunctive errors therein. Addressing these errors, we further propose heuristic dynamic integration constraint methods and rule-based static integration constraint methods to rectify errors and ultimately improve model performance. The entire methodology framework is termed as Transformer Doctor, designed for diagnosing and treating internal errors within transformers. Through a plethora of quantitative and qualitative experiments, it has been demonstrated that Transformer Doctor can effectively address internal errors in transformers, thereby enhancing model performance.