Vision-language models, like CLIP (Contrastive Language Image Pretraining), are becoming increasingly popular for a wide range of multimodal retrieval tasks. However, prior work has shown that large language and deep vision models can learn historical biases contained in their training sets, leading to perpetuation of stereotypes and potential downstream harm. In this work, we conduct a systematic analysis of the social biases that are present in CLIP, with a focus on the interaction between image and text modalities. We first propose a taxonomy of social biases called So-B-It, which contains 374 words categorized across ten types of bias. Each type can lead to societal harm if associated with a particular demographic group. Using this taxonomy, we examine images retrieved by CLIP from a facial image dataset using each word as part of a prompt. We find that CLIP frequently displays undesirable associations between harmful words and specific demographic groups, such as retrieving mostly pictures of Middle Eastern men when asked to retrieve images of a "terrorist". Finally, we conduct an analysis of the source of such biases, by showing that the same harmful stereotypes are also present in a large image-text dataset used to train CLIP models for examples of biases that we find. Our findings highlight the importance of evaluating and addressing bias in vision-language models, and suggest the need for transparency and fairness-aware curation of large pre-training datasets.

Supervised learning methods have been found to exhibit inductive biases favoring simpler features. When such features are spuriously correlated with the label, this can result in suboptimal performance on minority subgroups. Despite the growing popularity of methods which learn from unlabeled data, the extent to which these representations rely on spurious features for prediction is unclear. In this work, we explore the impact of spurious features on Self-Supervised Learning (SSL) for visual representation learning. We first empirically show that commonly used augmentations in SSL can cause undesired invariances in the image space, and illustrate this with a simple example. We further show that classical approaches in combating spurious correlations, such as dataset re-sampling during SSL, do not consistently lead to invariant representations. Motivated by these findings, we propose LateTVG to remove spurious information from these representations during pre-training, by regularizing later layers of the encoder via pruning. We find that our method produces representations which outperform the baselines on several benchmarks, without the need for group or label information during SSL.

Vision-language (VL) embedding models have been shown to encode biases present in their training data, such as societal biases that prescribe negative characteristics to members of various racial and gender identities. Due to their wide-spread adoption for various tasks ranging from few-shot classification to text-guided image generation, debiasing VL models is crucial. Debiasing approaches that fine-tune the VL model often suffer from catastrophic forgetting. On the other hand, fine-tuning-free methods typically utilize a ``one-size-fits-all" approach that assumes that correlation with the spurious attribute can be explained using a single linear direction across all possible inputs. In this work, we propose a nonlinear, fine-tuning-free approach for VL embedding model debiasing that tailors the debiasing operation to each unique input. This allows for a more flexible debiasing approach. Additionally, we do not require knowledge of the set of inputs a priori to inference time, making our method more appropriate for online tasks such as retrieval and text guided image generation.

Machine learning models can often fail on subgroups that are underrepresented during training. While dataset balancing can improve performance on underperforming groups, it requires access to training group annotations and can end up removing large portions of the dataset. In this paper, we introduce Data Debiasing with Datamodels (D3M), a debiasing approach which isolates and removes specific training examples that drive the model's failures on minority groups. Our approach enables us to efficiently train debiased classifiers while removing only a small number of examples, and does not require training group annotations or additional hyperparameter tuning.

BACKGROUND: The advent of high throughput sequencing has enabled researchers to systematically evaluate the genetic variations in cancer, identifying many cancer-associated genes. Although cancers in the same tissue are widely categorized in the same group, they demonstrate many differences concerning their mutational profiles. Hence, there is no definitive treatment for most cancer types. This reveals the importance of developing new pipelines to identify cancer-associated genes accurately and re-classify patients with similar mutational profiles. Classification of cancer patients with similar mutational profiles may help discover subtypes of cancer patients who might benefit from specific treatment types. RESULTS: In this study, we propose a new machine learning pipeline to identify protein-coding genes mutated in many samples to identify cancer subtypes. We apply our pipeline to 12,270 samples collected from the international cancer genome consortium, covering 19 cancer types. As a result, we identify 17 different cancer subtypes. Comprehensive phenotypic and genotypic analysis indicates distinguishable properties, including unique cancer-related signaling pathways. CONCLUSIONS: This new subtyping approach offers a novel opportunity for cancer drug development based on the mutational profile of patients. Additionally, we analyze the mutational signatures for samples in each subtype, which provides important insight into their active molecular mechanisms. Some of the pathways we identified in most subtypes, including the cell cycle and the Axon guidance pathways, are frequently observed in cancer disease. Interestingly, we also identified several mutated genes and different rates of mutation in multiple cancer subtypes. In addition, our study on "gene-motif" suggests the importance of considering both the context of the mutations and mutational processes in identifying cancer-associated genes. The source codes for our proposed clustering pipeline and analysis are publicly available at: https://github.com/bcb-sut/Pan-Cancer .