Foundation models (FMs) have achieved significant success across various tasks, leading to research on benchmarks for reasoning abilities. However, there is a lack of studies on FMs performance in exceptional scenarios, which we define as out-of-distribution (OOD) reasoning tasks. This paper is the first to address these cases, developing a novel dataset for evaluation of FMs across multiple modalities, including graphic novels, calligraphy, news articles, and lyrics. It includes tasks for instance classification, character recognition, token prediction, and text generation. The paper also proposes prompt engineering techniques like Chain-of-Thought (CoT) and CoT+Few-Shot to enhance performance. Validation of FMs using various methods revealed improvements. The code repository is accessible at: https://github.com/MLAI-Yonsei/ExceptionalBenchmark

Adapting a pre-trained foundation model on downstream tasks should ensure robustness against distribution shifts without the need to retrain the whole model. Although existing weight interpolation methods are simple yet effective, we argue their static nature limits downstream performance while achieving efficiency. In this work, we propose DaWin, a training-free dynamic weight interpolation method that leverages the entropy of individual models over each unlabeled test sample to assess model expertise, and compute per-sample interpolation coefficients dynamically. Unlike previous works that typically rely on additional training to learn such coefficients, our approach requires no training. Then, we propose a mixture modeling approach that greatly reduces inference overhead raised by dynamic interpolation. We validate DaWin on the large-scale visual recognition benchmarks, spanning 14 tasks across robust fine-tuning - ImageNet and derived five distribution shift benchmarks - and multi-task learning with eight classification tasks. Results demonstrate that DaWin achieves significant performance gain in considered settings, with minimal computational overhead. We further discuss DaWin's analytic behavior to explain its empirical success. The emergence of foundation models (Bommasani et al., 2021; Radford et al., 2021; Brown et al., 2020) has significantly lowered the barrier to deploying artificial intelligence solutions across a wide range of real-world problems. Leveraging the strong general knowledge acquired through large-scale pre-training, foundation models can be efficiently adapted for numerous tasks. However, recent studies have shown that while fine-tuning improves performance on specific downstream tasks, it may often undermine the model's generalizability and robustness (Wortsman et al., 2022b). For example, a model fine-tuned on ImageNet has better accuracy on in-distribution (ID) data yet may underperform in out-of-distribution (OOD) data such as ImageNet-A (Hendrycks et al., 2021b).

The large-scale pre-trained neural network has achieved notable success in enhancing performance for downstream tasks. Another promising approach for generalization is Bayesian Neural Network (BNN), which integrates Bayesian methods into neural network architectures, offering advantages such as Bayesian Model averaging (BMA) and uncertainty quantification. Despite these benefits, transfer learning for BNNs has not been widely investigated and shows limited improvement. We hypothesize that this issue arises from the inability to find flat minima, which is crucial for generalization performance. To address this, we evaluate the sharpness of BNNs in various settings, revealing their insufficiency in seeking flat minima and the influence of flatness on BMA performance. Therefore, we propose Sharpness-aware Bayesian Model Averaging (SA-BMA), a Bayesian-fitting flat posterior seeking optimizer integrated with Bayesian transfer learning. SA-BMA calculates the divergence between posteriors in the parameter space, aligning with the nature of BNNs, and serves as a generalized version of existing sharpness-aware optimizers. We validate that SA-BMA improves generalization performance in few-shot classification and distribution shift scenarios by ensuring flatness.

Approaches to improving multilingual language understanding often require multiple languages during the training phase, rely on complicated training techniques, and -- importantly -- struggle with significant performance gaps between high-resource and low-resource languages. We hypothesize that the performance gaps between languages are affected by linguistic gaps between those languages and provide a novel solution for robust multilingual language modeling by employing phonemic representations (specifically, using phonemes as input tokens to LMs rather than subwords). We present quantitative evidence from three cross-lingual tasks that demonstrate the effectiveness of phonemic representation, which is further justified by a theoretical analysis of the cross-lingual performance gap.

While fine-tuning unlocks the potential of a pre-trained model for a specific task, it compromises the model's ability to generalize to out-of-distribution (OOD) datasets. To mitigate this, robust fine-tuning aims to ensure performance on OOD datasets as well as on an in-distribution (ID) dataset for which the model is being tuned. However, another criterion for reliable machine learning (ML), confidence calibration, has been overlooked despite its increasing demand for real-world high-stakes ML applications (e.g., autonomous driving and medical diagnosis). For the first time, we raise concerns about the calibration of fine-tuned vision-language models (VLMs) under distribution shift by showing that naive fine-tuning and even state-of-the-art robust fine-tuning methods hurt the calibration of pre-trained VLMs, especially on OOD datasets. To address this issue, we provide a simple approach, called calibrated robust fine-tuning (CaRot), that incentivizes calibration and robustness on both ID and OOD datasets. Empirical results on ImageNet-1K distribution shift evaluation verify the effectiveness of our method.

Pre-trained multi-modal models, such as CLIP, provide transferable embeddings and show promising results in diverse applications. However, the analysis of learned multi-modal embeddings is relatively unexplored, and the embedding transferability can be improved. In this work, we observe that CLIP holds separated embedding subspaces for two different modalities, and then we investigate it through the lens of uniformity-alignment to measure the quality of learned representation. Both theoretically and empirically, we show that CLIP retains poor uniformity and alignment even after fine-tuning. Such a lack of alignment and uniformity might restrict the transferability and robustness of embeddings. To this end, we devise a new fine-tuning method for robust representation equipping better alignment and uniformity. First, we propose a Geodesic Multi-Modal Mixup that mixes the embeddings of image and text to generate hard negative samples on the hypersphere. Then, we fine-tune the model on hard negatives as well as original negatives and positives with contrastive loss. Based on the theoretical analysis about hardness guarantee and limiting behavior, we justify the use of our method. Extensive experiments on retrieval, calibration, few- or zero-shot classification (under distribution shift), embedding arithmetic, and image captioning further show that our method provides transferable representations, enabling robust model adaptation on diverse tasks. Code: https://github.com/changdaeoh/multimodal-mixup

We present a new supervised learning technique for the Variational AutoEncoder (VAE) that allows it to learn a causally disentangled representation and generate causally disentangled outcomes simultaneously. We call this approach Causally Disentangled Generation (CDG). CDG is a generative model that accurately decodes an output based on a causally disentangled representation. Our research demonstrates that adding supervised regularization to the encoder alone is insufficient for achieving a generative model with CDG, even for a simple task. Therefore, we explore the necessary and sufficient conditions for achieving CDG within a specific model. Additionally, we introduce a universal metric for evaluating the causal disentanglement of a generative model. Empirical results from both image and tabular datasets support our findings.

Machine learning algorithms have shown remarkable performance in diverse applications. However, it is still challenging to guarantee performance in distribution shifts when distributions of training and test datasets are different. There have been several approaches to improve the performance in distribution shift cases by learning invariant features across groups or domains. However, we observe that the previous works only learn invariant features partially. While the prior works focus on the limited invariant features, we first raise the importance of the sufficient invariant features. Since only training sets are given empirically, the learned partial invariant features from training sets might not be present in the test sets under distribution shift. Therefore, the performance improvement on distribution shifts might be limited. In this paper, we argue that learning sufficient invariant features from the training set is crucial for the distribution shift case. Concretely, we newly observe the connection between a) sufficient invariant features and b) flatness differences between groups or domains. Moreover, we propose a new algorithm, Adaptive Sharpness-aware Group Distributionally Robust Optimization (ASGDRO), to learn sufficient invariant features across domains or groups. ASGDRO learns sufficient invariant features by seeking common flat minima across all groups or domains. Therefore, ASGDRO improves the performance on diverse distribution shift cases. Besides, we provide a new simple dataset, Heterogeneous-CMNIST, to diagnose whether the various algorithms learn sufficient invariant features.

With the surge of large-scale pre-trained models (PTMs), fine-tuning these models to numerous downstream tasks becomes a crucial problem. Consequently, parameter efficient transfer learning (PETL) of large models has grasped huge attention. While recent PETL methods showcase impressive performance, they rely on optimistic assumptions: 1) the entire parameter set of a PTM is available, and 2) a sufficiently large memory capacity for the fine-tuning is equipped. However, in most real-world applications, PTMs are served as a black-box API or proprietary software without explicit parameter accessibility. Besides, it is hard to meet a large memory requirement for modern PTMs. In this work, we propose black-box visual prompting (BlackVIP), which efficiently adapts the PTMs without knowledge about model architectures and parameters. BlackVIP has two components; 1) Coordinator and 2) simultaneous perturbation stochastic approximation with gradient correction (SPSA-GC). The Coordinator designs input-dependent image-shaped visual prompts, which improves few-shot adaptation and robustness on distribution/location shift. SPSA-GC efficiently estimates the gradient of a target model to update Coordinator. Extensive experiments on 16 datasets demonstrate that BlackVIP enables robust adaptation to diverse domains without accessing PTMs' parameters, with minimal memory requirements. Code: \url{https://github.com/changdaeoh/BlackVIP}

Knowledge tracing plays a pivotal role in intelligent tutoring systems. This task aims to predict the probability of students answering correctly to specific questions. To do so, knowledge tracing systems should trace the knowledge state of the students by utilizing their problem-solving history and knowledge about the problems. Recent advances in knowledge tracing models have enabled better exploitation of problem solving history. However, knowledge about problems has not been studied, as well compared to students' answering histories. Knowledge tracing algorithms that incorporate knowledge directly are important to settings with limited data or cold starts. Therefore, we consider the problem of utilizing skill-to-skill relation to knowledge tracing. In this work, we introduce expert labeled skill-to-skill relationships. Moreover, we also provide novel methods to construct a knowledge-tracing model to leverage human experts' insight regarding relationships between skills. The results of an extensive experimental analysis show that our method outperformed a baseline Transformer model. Furthermore, we found that the extent of our model's superiority was greater in situations with limited data, which allows a smooth cold start of our model.