The increasing significance of large language and multimodal models in societal information processing has ignited debates on social safety and ethics. However, few studies have approached the analysis of these limitations from the comprehensive perspective of human and artificial intelligence system interactions. This study investigates biases and preferences when humans and large models are used as key links in communication. To achieve this, we design a multimodal dataset and three different experiments to evaluate generative models in their roles as producers and disseminators of information. Our main findings highlight that synthesized information is more likely to be incorporated into model training datasets and messaging than human-generated information. Additionally, large models, when acting as transmitters of information, tend to modify and lose specific content selectively. Conceptually, we present two realistic models of autophagic ("self-consumption") loops to account for the suppression of human-generated information in the exchange of information between humans and AI systems. We generalize the declining diversity of social information and the bottleneck in model performance caused by the above trends to the local optima of large models.
Video Corpus Moment Retrieval (VCMR) is a practical video retrieval task focused on identifying a specific moment within a vast corpus of untrimmed videos using the natural language query. Existing methods for VCMR typically rely on frame-aware video retrieval, calculating similarities between the query and video frames to rank videos based on maximum frame similarity.However, this approach overlooks the semantic structure embedded within the information between frames, namely, the event, a crucial element for human comprehension of videos. Motivated by this, we propose EventFormer, a model that explicitly utilizes events within videos as fundamental units for video retrieval. The model extracts event representations through event reasoning and hierarchical event encoding. The event reasoning module groups consecutive and visually similar frame representations into events, while the hierarchical event encoding encodes information at both the frame and event levels. We also introduce anchor multi-head self-attenion to encourage Transformer to capture the relevance of adjacent content in the video. The training of EventFormer is conducted by two-branch contrastive learning and dual optimization for two sub-tasks of VCMR. Extensive experiments on TVR, ANetCaps, and DiDeMo benchmarks show the effectiveness and efficiency of EventFormer in VCMR, achieving new state-of-the-art results. Additionally, the effectiveness of EventFormer is also validated on partially relevant video retrieval task.
While large language models (LMs) demonstrate remarkable performance, they encounter challenges in providing accurate responses when queried for information beyond their pre-trained memorization. Although augmenting them with relevant external information can mitigate these issues, failure to consider the necessity of retrieval may adversely affect overall performance. Previous research has primarily focused on examining how entities influence retrieval models and knowledge recall in LMs, leaving other aspects relatively unexplored. In this work, our goal is to offer a more detailed, fact-centric analysis by exploring the effects of combinations of entities and relations. To facilitate this, we construct a new question answering (QA) dataset called WiTQA (Wikipedia Triple Question Answers). This dataset includes questions about entities and relations of various popularity levels, each accompanied by a supporting passage. Our extensive experiments with diverse LMs and retrievers reveal when retrieval does not consistently enhance LMs from the viewpoints of fact-centric popularity.Confirming earlier findings, we observe that larger LMs excel in recalling popular facts. However, they notably encounter difficulty with infrequent entity-relation pairs compared to retrievers. Interestingly, they can effectively retain popular relations of less common entities. We demonstrate the efficacy of our finer-grained metric and insights through an adaptive retrieval system that selectively employs retrieval and recall based on the frequencies of entities and relations in the question.
Training image-based object detectors presents formidable challenges, as it entails not only the complexities of object detection but also the added intricacies of precisely localizing objects within potentially diverse and noisy environments. However, the collection of imagery itself can often be straightforward; for instance, cameras mounted in vehicles can effortlessly capture vast amounts of data in various real-world scenarios. In light of this, we introduce a groundbreaking method for training single-stage object detectors through unsupervised/self-supervised learning. Our state-of-the-art approach has the potential to revolutionize the labeling process, substantially reducing the time and cost associated with manual annotation. Furthermore, it paves the way for previously unattainable research opportunities, particularly for large, diverse, and challenging datasets lacking extensive labels. In contrast to prevalent unsupervised learning methods that primarily target classification tasks, our approach takes on the unique challenge of object detection. We pioneer the concept of intra-image contrastive learning alongside inter-image counterparts, enabling the acquisition of crucial location information essential for object detection. The method adeptly learns and represents this location information, yielding informative heatmaps. Our results showcase an outstanding accuracy of \textbf{89.2\%}, marking a significant breakthrough of approximately \textbf{15x} over random initialization in the realm of unsupervised object detection within the field of computer vision.
Deep neural networks have made great achievements in rainfall prediction.However, the current forecasting methods have certain limitations, such as with blurry generated images and incorrect spatial positions. To overcome these challenges, we propose a Two-stage Rainfall-Forecasting Diffusion Model (TRDM) aimed at improving the accuracy of long-term rainfall forecasts and addressing the imbalance in performance between temporal and spatial modeling. TRDM is a two-stage method for rainfall prediction tasks. The task of the first stage is to capture robust temporal information while preserving spatial information under low-resolution conditions. The task of the second stage is to reconstruct the low-resolution images generated in the first stage into high-resolution images. We demonstrate state-of-the-art results on the MRMS and Swedish radar datasets. Our project is open source and available on GitHub at: \href{https://github.com/clearlyzerolxd/TRDM}{https://github.com/clearlyzerolxd/TRDM}.
We present Polish Information Retrieval Benchmark (PIRB), a comprehensive evaluation framework encompassing 41 text information retrieval tasks for Polish. The benchmark incorporates existing datasets as well as 10 new, previously unpublished datasets covering diverse topics such as medicine, law, business, physics, and linguistics. We conduct an extensive evaluation of over 20 dense and sparse retrieval models, including the baseline models trained by us as well as other available Polish and multilingual methods. Finally, we introduce a three-step process for training highly effective language-specific retrievers, consisting of knowledge distillation, supervised fine-tuning, and building sparse-dense hybrid retrievers using a lightweight rescoring model. In order to validate our approach, we train new text encoders for Polish and compare their results with previously evaluated methods. Our dense models outperform the best solutions available to date, and the use of hybrid methods further improves their performance.
While there is no replacement for the learned expertise, devotion, and social benefits of a guide dog, there are cases in which a robot navigation assistant could be helpful for individuals with blindness or low vision (BLV). This study investigated the potential for an industrial agile robot to perform guided navigation tasks. We developed two interface prototypes that allowed for spatial information between a human-robot pair: a voice-based app and a flexible, responsive handle. The participants (n=21) completed simple navigation tasks and a post-study survey about the prototype functionality and their trust in the robot. All participants successfully completed the navigation tasks and demonstrated the interface prototypes were able to pass spatial information between the human and the robot. Future work will include expanding the voice-based app to allow the robot to communicate obstacles to the handler and adding haptic signals to the handle design.
A fair classifier should ensure the benefit of people from different groups, while the group information is often sensitive and unsuitable for model training. Therefore, learning a fair classifier but excluding sensitive attributes in the training dataset is important. In this paper, we study learning fair classifiers without implementing fair training algorithms to avoid possible leakage of sensitive information. Our theoretical analyses validate the possibility of this approach, that traditional training on a dataset with an appropriate distribution shift can reduce both the upper bound for fairness disparity and model generalization error, indicating that fairness and accuracy can be improved simultaneously with simply traditional training. We then propose a tractable solution to progressively shift the original training data during training by sampling influential data, where the sensitive attribute of new data is not accessed in sampling or used in training. Extensive experiments on real-world data demonstrate the effectiveness of our proposed algorithm.
Long-form generations from large language models (LLMs) contains a mix of factual and non-factual claims, making evaluating factuality difficult. To evaluate factual precision of long-form generations in a more fine-grained way, prior works propose to decompose long-form generations into multiple verifiable facts and verify those facts independently. The factuality of the generation is the proportion of verifiable facts among all the facts. Such methods assume that combining factual claims forms a factual paragraph. This paper shows that the assumption can be violated due to entity ambiguity. We show that LLMs can generate paragraphs that contain verifiable facts, but the facts are combined to form a non-factual paragraph due to entity ambiguity. We further reveal that existing factual precision metrics, including FActScore and citation recall, cannot properly evaluate the factuality of these non-factual paragraphs. To address this, we introduce an enhanced metric, D-FActScore, specifically designed for content with ambiguous entities. We evaluate the D-FActScores of people biographies generated with retrieval-augmented generation (RAG). We show that D-FActScore can better assess the factuality of paragraphs with entity ambiguity than FActScore. We also find that four widely used open-source LLMs tend to mix information of distinct entities to form non-factual paragraphs.
While large language models (LLMs) have achieved state-of-the-art performance on a wide range of medical question answering (QA) tasks, they still face challenges with hallucinations and outdated knowledge. Retrieval-augmented generation (RAG) is a promising solution and has been widely adopted. However, a RAG system can involve multiple flexible components, and there is a lack of best practices regarding the optimal RAG setting for various medical purposes. To systematically evaluate such systems, we propose the Medical Information Retrieval-Augmented Generation Evaluation (MIRAGE), a first-of-its-kind benchmark including 7,663 questions from five medical QA datasets. Using MIRAGE, we conducted large-scale experiments with over 1.8 trillion prompt tokens on 41 combinations of different corpora, retrievers, and backbone LLMs through the MedRAG toolkit introduced in this work. Overall, MedRAG improves the accuracy of six different LLMs by up to 18% over chain-of-thought prompting, elevating the performance of GPT-3.5 and Mixtral to GPT-4-level. Our results show that the combination of various medical corpora and retrievers achieves the best performance. In addition, we discovered a log-linear scaling property and the "lost-in-the-middle" effects in medical RAG. We believe our comprehensive evaluations can serve as practical guidelines for implementing RAG systems for medicine.