Significant scientific discoveries have driven the progress of human civilisation. The explosion of scientific literature and data has created information barriers across disciplines that have slowed the pace of scientific discovery. Large Language Models (LLMs) hold a wealth of global and interdisciplinary knowledge that promises to break down these information barriers and foster a new wave of scientific discovery. However, the potential of LLMs for scientific discovery has not been formally explored. In this paper, we start from investigating whether LLMs can propose scientific hypotheses. To this end, we construct a dataset consist of background knowledge and hypothesis pairs from biomedical literature. The dataset is divided into training, seen, and unseen test sets based on the publication date to control visibility. We subsequently evaluate the hypothesis generation capabilities of various top-tier instructed models in zero-shot, few-shot, and fine-tuning settings, including both closed and open-source LLMs. Additionally, we introduce an LLM-based multi-agent cooperative framework with different role designs and external tools to enhance the capabilities related to generating hypotheses. We also design four metrics through a comprehensive review to evaluate the generated hypotheses for both ChatGPT-based and human evaluations. Through experiments and analyses, we arrive at the following findings: 1) LLMs surprisingly generate untrained yet validated hypotheses from testing literature. 2) Increasing uncertainty facilitates candidate generation, potentially enhancing zero-shot hypothesis generation capabilities. These findings strongly support the potential of LLMs as catalysts for new scientific discoveries and guide further exploration.
We study stereotypes embedded within one of the most popular text-to-image generators: Stable Diffusion. We examine what stereotypes of gender and nationality/continental identity does Stable Diffusion display in the absence of such information i.e. what gender and nationality/continental identity is assigned to `a person', or to `a person from Asia'. Using vision-language model CLIP's cosine similarity to compare images generated by CLIP-based Stable Diffusion v2.1 verified by manual examination, we chronicle results from 136 prompts (50 results/prompt) of front-facing images of persons from 6 different continents, 27 nationalities and 3 genders. We observe how Stable Diffusion outputs of `a person' without any additional gender/nationality information correspond closest to images of men and least with persons of nonbinary gender, and to persons from Europe/North America over Africa/Asia, pointing towards Stable Diffusion having a concerning representation of personhood to be a European/North American man. We also show continental stereotypes and resultant harms e.g. a person from Oceania is deemed to be Australian/New Zealander over Papua New Guinean, pointing to the erasure of Indigenous Oceanic peoples, who form a majority over descendants of colonizers both in Papua New Guinea and in Oceania overall. Finally, we unexpectedly observe a pattern of oversexualization of women, specifically Latin American, Mexican, Indian and Egyptian women relative to other nationalities, measured through an NSFW detector. This demonstrates how Stable Diffusion perpetuates Western fetishization of women of color through objectification in media, which if left unchecked will amplify this stereotypical representation. Image datasets are made publicly available.
Recommendation models trained on the user feedback collected from deployed recommendation systems are commonly biased. User feedback is considerably affected by the exposure mechanism, as users only provide feedback on the items exposed to them and passively ignore the unexposed items, thus producing numerous false negative samples. Inevitably, biases caused by such user feedback are inherited by new models and amplified via feedback loops. Moreover, the presence of false negative samples makes negative sampling difficult and introduces spurious information in the user preference modeling process of the model. Recent work has investigated the negative impact of feedback loops and unknown exposure mechanisms on recommendation quality and user experience, essentially treating them as independent factors and ignoring their cross-effects. To address these issues, we deeply analyze the data exposure mechanism from the perspective of data iteration and feedback loops with the Missing Not At Random (\textbf{MNAR}) assumption, theoretically demonstrating the existence of an available stabilization factor in the transformation of the exposure mechanism under the feedback loops. We further propose Dynamic Personalized Ranking (\textbf{DPR}), an unbiased algorithm that uses dynamic re-weighting to mitigate the cross-effects of exposure mechanisms and feedback loops without additional information. Furthermore, we design a plugin named Universal Anti-False Negative (\textbf{UFN}) to mitigate the negative impact of the false negative problem. We demonstrate theoretically that our approach mitigates the negative effects of feedback loops and unknown exposure mechanisms. Experimental results on real-world datasets demonstrate that models using DPR can better handle bias accumulation and the universality of UFN in mainstream loss methods.
Within the context of Graph Signal Processing (GSP), Graph Learning (GL) is concerned with the inference of the graph's underlying structure from nodal observations. However, real-world data often contains diverse information, necessitating the simultaneous clustering and learning of multiple graphs. In practical applications, valuable node-specific covariates, represented as kernels, have been underutilized by existing graph signal clustering methods. In this letter, we propose a new framework, named Kernel-based joint Multiple GL and clustering of graph signals (KMGL), that leverages a multi-convex optimization approach. This allows us to integrate node-side information, construct low-pass filters, and efficiently solve the optimization problem. The experiments demonstrate that KMGL significantly enhances the robustness of GL and clustering, particularly in scenarios with high noise levels and a substantial number of clusters. These findings underscore the potential of KMGL for improving the performance of GSP methods in diverse, real-world applications.
Despite the dramatic progress in Large Language Model (LLM) development, LLMs often provide seemingly plausible but not factual information, often referred to as hallucinations. Retrieval-augmented LLMs provide a non-parametric approach to solve these issues by retrieving relevant information from external data sources and augment the training process. These models help to trace evidence from an externally provided knowledge base allowing the model predictions to be better interpreted and verified. In this work, we critically evaluate these models in their ability to perform in scientific document reasoning tasks. To this end, we tuned multiple such model variants with science-focused instructions and evaluated them on a scientific document reasoning benchmark for the usefulness of the retrieved document passages. Our findings suggest that models justify predictions in science tasks with fabricated evidence and leveraging scientific corpus as pretraining data does not alleviate the risk of evidence fabrication.
Accurate detection of inundated water extents during flooding events is crucial in emergency response decisions and aids in recovery efforts. Satellite Remote Sensing data provides a global framework for detecting flooding extents. Specifically, Sentinel-1 C-Band Synthetic Aperture Radar (SAR) imagery has proven to be useful in detecting water bodies due to low backscatter of water features in both co-polarized and cross-polarized SAR imagery. However, increased backscatter can be observed in certain flooded regions such as presence of infrastructure and trees - rendering simple methods such as pixel intensity thresholding and time-series differencing inadequate. Machine Learning techniques has been leveraged to precisely capture flood extents in flooded areas with bumps in backscatter but needs high amounts of labelled data to work desirably. Hence, we created a labeled known water body extent and flooded area extents during known flooding events covering about 36,000 sq. kilometers of regions within mainland U.S and Bangladesh. Further, We also leveraged citizen science by open-sourcing the dataset and hosting an open competition based on the dataset to rapidly prototype flood extent detection using community generated models. In this paper we present the information about the dataset, the data processing pipeline, a baseline model and the details about the competition, along with discussion on winning approaches. We believe the dataset adds to already existing datasets based on Sentinel-1C SAR data and leads to more robust modeling of flood extents. We also hope the results from the competition pushes the research in flood extent detection further.
US corporations regularly spend millions of dollars reviewing electronically-stored documents in legal matters. Recently, attorneys apply text classification to efficiently cull massive volumes of data to identify responsive documents for use in these matters. While text classification is regularly used to reduce the discovery costs of legal matters, it also faces a perception challenge: amongst lawyers, this technology is sometimes looked upon as a "black box". Put simply, no extra information is provided for attorneys to understand why documents are classified as responsive. In recent years, explainable machine learning has emerged as an active research area. In an explainable machine learning system, predictions or decisions made by a machine learning model are human understandable. In legal 'document review' scenarios, a document is responsive, because one or more of its small text snippets are deemed responsive. In these scenarios, if these responsive snippets can be located, then attorneys could easily evaluate the model's document classification decisions - this is especially important in the field of responsible AI. Our prior research identified that predictive models created using annotated training text snippets improved the precision of a model when compared to a model created using all of a set of documents' text as training. While interesting, manually annotating training text snippets is not generally practical during a legal document review. However, small increases in precision can drastically decrease the cost of large document reviews. Automating the identification of training text snippets without human review could then make the application of training text snippet-based models a practical approach.
In this work, we study integrated sensing and communication (ISAC) networks with the aim of effectively balancing sensing and communication (S&C) performance at the network level. Focusing on monostatic sensing, the tool of stochastic geometry is exploited to capture the S&C performance, which facilitates us to illuminate key cooperative dependencies in the ISAC network and optimize key network-level parameters. Based on the derived tractable expression of area spectral efficiency (ASE), we formulate the optimization problem to maximize the network performance from the view point of two joint S&C metrics. Towards this end, we further jointly optimize the cooperative BS cluster sizes for S&C and the serving/probing numbers of users/targets to achieve a flexible tradeoff between S&C at the network level. It is verified that interference nulling can effectively improve the average data rate and radar information rate. Surprisingly, the optimal communication tradeoff for the case of the ASE maximization tends to employ all spacial resources towards multiplexing and diversity gain, without interference nulling. By contrast, for the sensing objectives, resource allocation tends to eliminate certain interference especially when the antenna resources are sufficient, because the inter-cell interference becomes a more dominant factor affecting sensing performance. Furthermore, we prove that the ratio of the optimal number of users and the number of transmit antennas is a constant value when the communication performance is optimal. Simulation results demonstrate that the proposed cooperative ISAC scheme achieves a substantial gain in S&C performance at the network level.
Defects in laser powder bed fusion (L-PBF) parts often result from the meso-scale dynamics of the molten alloy near the laser, known as the melt pool. For instance, the melt pool can directly contribute to the formation of undesirable porosity, residual stress, and surface roughness in the final part. Experimental in-situ monitoring of the three-dimensional melt pool physical fields is challenging, due to the short length and time scales involved in the process. Multi-physics simulation methods can describe the three-dimensional dynamics of the melt pool, but are computationally expensive at the mesh refinement required for accurate predictions of complex effects, such as the formation of keyhole porosity. Therefore, in this work, we develop a generative deep learning model based on the probabilistic diffusion framework to map low-fidelity, coarse-grained simulation information to the high-fidelity counterpart. By doing so, we bypass the computational expense of conducting multiple high-fidelity simulations for analysis by instead upscaling lightweight coarse mesh simulations. Specifically, we implement a 2-D diffusion model to spatially upscale cross-sections of the coarsely simulated melt pool to their high-fidelity equivalent. We demonstrate the preservation of key metrics of the melting process between the ground truth simulation data and the diffusion model output, such as the temperature field, the melt pool dimensions and the variability of the keyhole vapor cavity. Specifically, we predict the melt pool depth within 3 $\mu m$ based on low-fidelity input data 4$\times$ coarser than the high-fidelity simulations, reducing analysis time by two orders of magnitude.
Target volume contouring for radiation therapy is considered significantly more challenging than the normal organ segmentation tasks as it necessitates the utilization of both image and text-based clinical information. Inspired by the recent advancement of large language models (LLMs) that can facilitate the integration of the textural information and images, here we present a novel LLM-driven multi-modal AI that utilizes the clinical text information and is applicable to the challenging task of target volume contouring for radiation therapy, and validate it within the context of breast cancer radiation therapy target volume contouring. Using external validation and data-insufficient environments, which attributes highly conducive to real-world applications, we demonstrate that the proposed model exhibits markedly improved performance compared to conventional vision-only AI models, particularly exhibiting robust generalization performance and data-efficiency. To our best knowledge, this is the first LLM-driven multimodal AI model that integrates the clinical text information into target volume delineation for radiation oncology.