Restoring Calibration for Aligned Large Language Models: A Calibration-Aware Fine-Tuning Approach

One of the key technologies for the success of Large Language Models (LLMs) is preference alignment. However, a notable side effect of preference alignment is poor calibration: while the pre-trained models are typically well-calibrated, LLMs tend to become poorly calibrated after alignment with human preferences. In this paper, we investigate why preference alignment affects calibration and how to address this issue. For the first question, we observe that the preference collapse issue in alignment undesirably generalizes to the calibration scenario, causing LLMs to exhibit overconfidence and poor calibration. To address this, we demonstrate the importance of fine-tuning with domain-specific knowledge to alleviate the overconfidence issue. To further analyze whether this affects the model's performance, we categorize models into two regimes: calibratable and non-calibratable, defined by bounds of Expected Calibration Error (ECE). In the calibratable regime, we propose a calibration-aware fine-tuning approach to achieve proper calibration without compromising LLMs' performance. However, as models are further fine-tuned for better performance, they enter the non-calibratable regime. For this case, we develop an EM-algorithm-based ECE regularization for the fine-tuning loss to maintain low calibration error. Extensive experiments validate the effectiveness of the proposed methods.

ICAFS: Inter-Client-Aware Feature Selection for Vertical Federated Learning

Vertical federated learning (VFL) enables a paradigm for vertically partitioned data across clients to collaboratively train machine learning models. Feature selection (FS) plays a crucial role in Vertical Federated Learning (VFL) due to the unique nature that data are distributed across multiple clients. In VFL, different clients possess distinct subsets of features for overlapping data samples, making the process of identifying and selecting the most relevant features a complex yet essential task. Previous FS efforts have primarily revolved around intra-client feature selection, overlooking vital feature interaction across clients, leading to subpar model outcomes. We introduce ICAFS, a novel multi-stage ensemble approach for effective FS in VFL by considering inter-client interactions. By employing conditional feature synthesis alongside multiple learnable feature selectors, ICAFS facilitates ensemble FS over these selectors using synthetic embeddings. This method bypasses the limitations of private gradient sharing and allows for model training using real data with refined embeddings. Experiments on multiple real-world datasets demonstrate that ICAFS surpasses current state-of-the-art methods in prediction accuracy.

Fine-Tuning Linear Layers Only Is a Simple yet Effective Way for Task Arithmetic

Task arithmetic has recently emerged as a cost-effective and scalable approach to edit pre-trained models directly in weight space, by adding the fine-tuned weights of different tasks. The performance has been further improved by a linear property which is illustrated by weight disentanglement. Yet, conventional linearization methods (e.g., NTK linearization) not only double the time and training cost but also have a disadvantage on single-task performance. We propose a simple yet effective and efficient method that only fine-tunes linear layers, which improves weight disentanglement and efficiency simultaneously. Specifically, our study reveals that only fine-tuning the linear layers in the attention modules makes the whole model occur in a linear regime, significantly improving weight disentanglement. To further understand how our method improves the disentanglement of task arithmetic, we present a comprehensive study of task arithmetic by differentiating the role of representation model and task-specific model. In particular, we find that the representation model plays an important role in improving weight disentanglement whereas the task-specific models such as the classification heads can degenerate the weight disentanglement performance. Overall, our work uncovers novel insights into the fundamental mechanisms of task arithmetic and offers a more reliable and effective approach to editing pre-trained models.