Comprehensive Benchmark Research Articles

RS-LLaVA: A Large Vision-Language Model for Joint Captioning and Question Answering in Remote Sensing Imagery

In this paper, we delve into the innovative application of large language models (LLMs) and their extension, large vision-language models (LVLMs), in the field of remote sensing (RS) image analysis. We particularly emphasize their multi-tasking potential with a focus on image captioning and visual question answering (VQA). In particular, we introduce an improved version of the Large Language and Vision Assistant Model (LLaVA), specifically adapted for RS imagery through a low-rank adaptation approach. To evaluate the model performance, we create the RS-instructions dataset, a comprehensive benchmark dataset that integrates four diverse single-task datasets related to captioning and VQA. The experimental results confirm the model’s effectiveness, marking a step forward toward the development of efficient multi-task models for RS image analysis.

Revisiting face detection: Supercharging Viola-Jones with particle swarm optimization for enhanced performance

In recent years, face detection has emerged as a prominent research field within Computer Vision (CV) and Deep Learning. Detecting faces in images and video sequences remains a challenging task due to various factors such as pose variation, varying illumination, occlusion, and scale differences. Despite the development of numerous face detection algorithms in deep learning, the Viola-Jones algorithm, with its simple yet effective approach, continues to be widely used in real-time camera applications. The conventional Viola-Jones algorithm employs AdaBoost for classifying faces in images and videos. The challenge lies in working with cluttered real-time facial images. AdaBoost needs to search through all possible thresholds for all samples to find the minimum training error when receiving features from Haar-like detectors. Therefore, this exhaustive search consumes significant time to discover the best threshold values and optimize feature selection to build an efficient classifier for face detection. In this paper, we propose enhancing the conventional Viola-Jones algorithm by incorporating Particle Swarm Optimization (PSO) to improve its predictive accuracy, particularly in complex face images. We leverage PSO in two key areas within the Viola-Jones framework. Firstly, PSO is employed to dynamically select optimal threshold values for feature selection, thereby improving computational efficiency. Secondly, we adapt the feature selection process using AdaBoost within the Viola-Jones algorithm, integrating PSO to identify the most discriminative features for constructing a robust classifier. Our approach significantly reduces the feature selection process time and search complexity compared to the traditional algorithm, particularly in challenging environments. We evaluated our proposed method on a comprehensive face detection benchmark dataset, achieving impressive results, including an average true positive rate of 98.73% and a 2.1% higher average prediction accuracy when compared against both the conventional Viola-Jones approach and contemporary state-of-the-art methods.

Modeling single cell trajectory using forward-backward stochastic differential equations.

Recent advances in single-cell sequencing technology have provided opportunities for mathematical modeling of dynamic developmental processes at the single-cell level, such as inferring developmental trajectories. Optimal transport has emerged as a promising theoretical framework for this task by computing pairings between cells from different time points. However, optimal transport methods have limitations in capturing nonlinear trajectories, as they are static and can only infer linear paths between endpoints. In contrast, stochastic differential equations (SDEs) offer a dynamic and flexible approach that can model non-linear trajectories, including the shape of the path. Nevertheless, existing SDE methods often rely on numerical approximations that can lead to inaccurate inferences, deviating from true trajectories. To address this challenge, we propose a novel approach combining forward-backward stochastic differential equations (FBSDE) with a refined approximation procedure. Our FBSDE model integrates the forward and backward movements of two SDEs in time, aiming to capture the underlying dynamics of single-cell developmental trajectories. Through comprehensive benchmarking on multiple scRNA-seq datasets, we demonstrate the superior performance of FBSDE compared to other methods, highlighting its efficacy in accurately inferring developmental trajectories.

Abstract P19: Comprehensive Benchmark of Somatic Mutation Callers on Cell-Free DNA Next-Generation Sequencing Data in Cancer Liquid Biopsy

Abstract Circulating tumour DNA (ctDNA) has emerged as a rich source of biomarkers in precision oncology. However, ctDNA sequence data is often noisier than that derived from standard tissue biopsies, posing significant challenges for accurate cancer mutation detection. Besides, there is a lack of standardised and unbiased benchmark datasets to evaluate and compare performance of variant callers on ctDNA. To address this issue, we present a novel benchmarking approach and comprehensive benchmark of cancer somatic mutation callers on cell-free DNA (cfDNA) Next Generation Sequencing (NGS) data. To create a large-scale benchmark dataset, we developed a new methodology using real colorectal cancer patients plasma samples exhibiting either ultra-high and an ultra-low levels of ctDNA. This enabled the generation of dilution series of deep Whole Genome Sequencing (WGS, 150x) and the creation of high-quality consensus ground truth calls. We evaluated the performance of 9 SNV callers and 7 short INDEL callers frequently used in academic research, including 2 callers that were specifically designed for cfDNA. We further validated our findings at higher depth of coverage using deep Whole Exome Sequencing (WES, >2,000x) data. We further investigated each caller's features to develop practical guidelines for users depending on their application setting: unguided analysis focusing on overall accuracy vs. guided analysis focusing on sensitivity. Citation Format: Hanaé Carrié, Ngak Leng Sim, Pui Mun Wong, Anna Gan, Yi Ting Lau, Polly Poon, Saranya Thangaraju, Iain Tan, Yoon Sim Yap, Limsoon Wong, Anders Skanderup. Comprehensive Benchmark of Somatic Mutation Callers on Cell-Free DNA Next-Generation Sequencing Data in Cancer Liquid Biopsy [abstract]. In: Proceedings of Frontiers in Cancer Science; 2023 Nov 6-8; Singapore. Philadelphia (PA): AACR; Cancer Res 2024;84(8_Suppl):Abstract nr P19.

Multi-Context Point Cloud Dataset and Machine Learning for Railway Semantic Segmentation

Railway scene understanding is crucial for various applications, including autonomous trains, digital twining, and infrastructure change monitoring. However, the development of the latter is constrained by the lack of annotated datasets and limitations of existing algorithms. To address this challenge, we present Rail3D, the first comprehensive dataset for semantic segmentation in railway environments with a comparative analysis. Rail3D encompasses three distinct railway contexts from Hungary, France, and Belgium, capturing a wide range of railway assets and conditions. With over 288 million annotated points, Rail3D surpasses existing datasets in size and diversity, enabling the training of generalizable machine learning models. We conducted a generic classification with nine universal classes (Ground, Vegetation, Rail, Poles, Wires, Signals, Fence, Installation, and Building) and evaluated the performance of three state-of-the-art models: KPConv (Kernel Point Convolution), LightGBM, and Random Forest. The best performing model, a fine-tuned KPConv, achieved a mean Intersection over Union (mIoU) of 86%. While the LightGBM-based method achieved a mIoU of 71%, outperforming Random Forest. This study will benefit infrastructure experts and railway researchers by providing a comprehensive dataset and benchmarks for 3D semantic segmentation. The data and code are publicly available for France and Hungary, with continuous updates based on user feedback.

A comprehensive benchmark of graph-based genetic variant genotyping algorithms on plant genomes for creating an accurate ensemble pipeline.

Although sequencing technologies have boosted the measurement of the genomic diversity of plant crops, it remains challenging to accurately genotype millions of genetic variants, especially structural variations, with only short reads. In recent years, many graph-based variation genotyping methods have been developed to address this issue and tested for human genomes. However, their performance in plant genomes remains largely elusive. Furthermore, pipelines integrating the advantages of current genotyping methods might be required, considering the different complexity of plant genomes. Here we comprehensively evaluate eight such genotypers in different scenarios in terms of variant type and size, sequencing parameters, genomic context, and complexity, as well as graph size, using both simulated and real data sets from representative plant genomes. Our evaluation reveals that there are still great challenges to applying existing methods to plants, such as excessive repeats and variants or high resource consumption. Therefore, we propose a pipeline called Ensemble Variant Genotyper (EVG) that can achieve better genotyping performance in almost all experimental scenarios and comparably higher genotyping recall and precision even using 5× reads. Furthermore, we demonstrate that EVG is more robust with an increasing number of graphed genomes, especially for insertions and deletions. Our study will provide new insights into the development and application of graph-based genotyping algorithms. We conclude that EVG provides an accurate, unbiased, and cost-effective way for genotyping both small and large variations and will be potentially used in population-scale genotyping for large, repetitive, and heterozygous plant genomes.

Radiative Heat Transfer Measurements of Titan Atmospheric Entry in a Shock Tube

Measurements were performed in the T6 Stalker facility operating in Aluminum Shock Tube mode for conditions relevant to Titan entry. Spatially and spectrally resolved radiation emitted from a high-temperature test gas behind a normal shock was recorded by means of emission spectroscopy. For Titan atmospheric entry, the main radiator of interest is cyano radical, formed in the nonequilibrium region behind the shock. The tests reported in this work measured radiation at velocities from 3.1 to 8.5 km/s and freestream pressures of 13, 20, and 133 Pa at a nominal composition of 98% N2 and 2% CH4. These shock layer radiation experiments employed an optical emission spectroscopy system on each side of the facility to allow two spectral regions to be measured simultaneously for each test, covering the spectral range of 200–900 nm. The present work provides a new, comprehensive benchmark set of data relevant to Titan entry studies.

Synergetic Assessment of Quality and Aesthetic: Approach and Comprehensive Benchmark Dataset

Synergetic Assessment of Quality and Aesthetic: Approach and Comprehensive Benchmark Dataset

ConvNeXt-MHC: improving MHC-peptide affinity prediction by structure-derived degenerate coding and the ConvNeXt model.

Peptide binding to major histocompatibility complex (MHC) proteins plays a critical role in T-cell recognition and the specificity of the immune response. Experimental validation such peptides is extremely resource-intensive. As a result, accurate computational prediction of binding peptides is highly important, particularly in the context of cancer immunotherapy applications, such as the identification of neoantigens. In recent years, there is a significant need to continually improve the existing prediction methods to meet the demands of this field. We developed ConvNeXt-MHC, a method for predicting MHC-I-peptide binding affinity. It introduces a degenerate encoding approach to enhance well-established panspecific methods and integrates transfer learning and semi-supervised learning methods into the cutting-edge deep learning framework ConvNeXt. Comprehensive benchmark results demonstrate that ConvNeXt-MHC outperforms state-of-the-art methods in terms of accuracy. We expect that ConvNeXt-MHC will help us foster new discoveries in the field of immunoinformatics in the distant future. We constructed a user-friendly website at http://www.combio-lezhang.online/predict/, where users can access our data and application.

Extend the benchmarking indel set by manual review using the individual cell line sequencing data from the Sequencing Quality Control 2 (SEQC2) project

Accurate indel calling plays an important role in precision medicine. A benchmarking indel set is essential for thoroughly evaluating the indel calling performance of bioinformatics pipelines. A reference sample with a set of known-positive variants was developed in the FDA-led Sequencing Quality Control Phase 2 (SEQC2) project, but the known indels in the known-positive set were limited. This project sought to provide an enriched set of known indels that would be more translationally relevant by focusing on additional cancer related regions. A thorough manual review process completed by 42 reviewers, two advisors, and a judging panel of three researchers significantly enriched the known indel set by an additional 516 indels. The extended benchmarking indel set has a large range of variant allele frequencies (VAFs), with 87% of them having a VAF below 20% in reference Sample A. The reference Sample A and the indel set can be used for comprehensive benchmarking of indel calling across a wider range of VAF values in the lower range. Indel length was also variable, but the majority were under 10 base pairs (bps). Most of the indels were within coding regions, with the remainder in the gene regulatory regions. Although high confidence can be derived from the robust study design and meticulous human review, this extensive indel set has not undergone orthogonal validation. The extended benchmarking indel set, along with the indels in the previously published known-positive set, was the truth set used to benchmark indel calling pipelines in a community challenge hosted on the precisionFDA platform. This benchmarking indel set and reference samples can be utilized for a comprehensive evaluation of indel calling pipelines. Additionally, the insights and solutions obtained during the manual review process can aid in improving the performance of these pipelines.

Mono3DVG: 3D Visual Grounding in Monocular Images

We introduce a novel task of 3D visual grounding in monocular RGB images using language descriptions with both appearance and geometry information. Specifically, we build a large-scale dataset, Mono3DRefer, which contains 3D object targets with their corresponding geometric text descriptions, generated by ChatGPT and refined manually. To foster this task, we propose Mono3DVG-TR, an end-to-end transformer-based network, which takes advantage of both the appearance and geometry information in text embeddings for multi-modal learning and 3D object localization. Depth predictor is designed to explicitly learn geometry features. The dual text-guided adapter is proposed to refine multiscale visual and geometry features of the referred object. Based on depth-text-visual stacking attention, the decoder fuses object-level geometric cues and visual appearance into a learnable query. Comprehensive benchmarks and some insightful analyses are provided for Mono3DVG. Extensive comparisons and ablation studies show that our method significantly outperforms all baselines. The dataset and code will be released.

Toward Open-Set Human Object Interaction Detection

This work is oriented toward the task of open-set Human Object Interaction (HOI) detection. The challenge lies in identifying completely new, out-of-domain relationships, as opposed to in-domain ones which have seen improvements in zero-shot HOI detection. To address this challenge, we introduce a simple Disentangled HOI Detection (DHD) model for detecting novel relationships by integrating an open-set object detector with a Visual Language Model (VLM). We utilize a disentangled image-text contrastive learning metric for training and connect the bottom-up visual features to text embeddings through lightweight unary and pair-wise adapters. Our model can benefit from the open-set object detector and the VLM to detect novel action categories and combine actions with novel object categories. We further present the VG-HOI dataset, a comprehensive benchmark with over 17k HOI relationships for open-set scenarios. Experimental results show that our model can detect unknown action classes and combine unknown object classes. Furthermore, it can generalize to over 17k HOI classes while being trained on just 600 HOI classes.

SciEval: A Multi-Level Large Language Model Evaluation Benchmark for Scientific Research

Recently, there has been growing interest in using Large Language Models (LLMs) for scientific research. Numerous benchmarks have been proposed to evaluate the ability of LLMs for scientific research. However, current benchmarks are mostly based on pre-collected objective questions. This design suffers from data leakage problem and lacks the evaluation of subjective Q/A ability. In this paper, we propose SciEval, a comprehensive and multi-disciplinary evaluation benchmark to address these issues. Based on Bloom's taxonomy, SciEval covers four dimensions to systematically evaluate scientific research ability. In particular, we design a "dynamic" subset based on scientific principles to prevent evaluation from potential data leakage. Both objective and subjective questions are included in SciEval. These characteristics make SciEval a more effective benchmark for scientific research ability evaluation of LLMs. Comprehensive experiments on most advanced LLMs show that, although GPT-4 achieves SOTA performance compared to other LLMs, there is still substantial room for improvement, especially for dynamic questions. The codes and data are publicly available on https://github.com/OpenDFM/SciEval.

BLIVA: A Simple Multimodal LLM for Better Handling of Text-Rich Visual Questions

Vision Language Models (VLMs), which extend Large Language Models (LLM) by incorporating visual understanding capability, have demonstrated significant advancements in addressing open-ended visual question-answering (VQA) tasks. However, these models cannot accurately interpret images infused with text, a common occurrence in real-world scenarios. Standard procedures for extracting information from images often involve learning a fixed set of query embeddings. These embeddings are designed to encapsulate image contexts and are later used as soft prompt inputs in LLMs. Yet, this process is limited to the token count, potentially curtailing the recognition of scenes with text-rich context. To improve upon them, the present study introduces BLIVA: an augmented version of InstructBLIP with Visual Assistant. BLIVA incorporates the query embeddings from InstructBLIP and also directly projects encoded patch embeddings into the LLM, a technique inspired by LLaVA. This approach assists the model to capture intricate details potentially missed during the query decoding process. Empirical evidence demonstrates that our model, BLIVA, significantly enhances performance in processing text-rich VQA benchmarks (up to 17.76% in OCR-VQA benchmark) and in undertaking general (not particularly text-rich) VQA benchmarks (up to 7.9% in Visual Spatial Reasoning benchmark), and achieved 17.72% overall improvement in a comprehensive multimodal LLM benchmark (MME), comparing to our baseline InstructBLIP. BLIVA demonstrates significant capability in decoding real-world images, irrespective of text presence. To demonstrate the broad industry applications enabled by BLIVA, we evaluate the model using a new dataset comprising YouTube thumbnails paired with question-answer sets across 11 diverse categories. For researchers interested in further exploration, our code and models are freely accessible at https://github.com/mlpc-ucsd/BLIVA.

Region-Disentangled Diffusion Model for High-Fidelity PPG-to-ECG Translation

The high prevalence of cardiovascular diseases (CVDs) calls for accessible and cost-effective continuous cardiac monitoring tools. Despite Electrocardiography (ECG) being the gold standard, continuous monitoring remains a challenge, leading to the exploration of Photoplethysmography (PPG), a promising but more basic alternative available in consumer wearables. This notion has recently spurred interest in translating PPG to ECG signals. In this work, we introduce Region-Disentangled Diffusion Model (RDDM), a novel diffusion model designed to capture the complex temporal dynamics of ECG. Traditional Diffusion models like Denoising Diffusion Probabilistic Models (DDPM) face challenges in capturing such nuances due to the indiscriminate noise addition process across the entire signal. Our proposed RDDM overcomes such limitations by incorporating a novel forward process that selectively adds noise to specific regions of interest (ROI) such as QRS complex in ECG signals, and a reverse process that disentangles the denoising of ROI and non-ROI regions. Quantitative experiments demonstrate that RDDM can generate high-fidelity ECG from PPG in as few as 10 diffusion steps, making it highly effective and computationally efficient. Additionally, to rigorously validate the usefulness of the generated ECG signals, we introduce CardioBench, a comprehensive evaluation benchmark for a variety of cardiac-related tasks including heart rate and blood pressure estimation, stress classification, and the detection of atrial fibrillation and diabetes. Our thorough experiments show that RDDM achieves state-of-the-art performance on CardioBench. To the best of our knowledge, RDDM is the first diffusion model for cross-modal signal-to-signal translation in the bio-signal domain.

BirdCollect: A Comprehensive Benchmark for Analyzing Dense Bird Flock Attributes

Automatic recognition of bird behavior from long-term, un controlled outdoor imagery can contribute to conservation efforts by enabling large-scale monitoring of bird populations. Current techniques in AI-based wildlife monitoring have focused on short-term tracking and monitoring birds individually rather than in species-rich flocks. We present Bird-Collect, a comprehensive benchmark dataset for monitoring dense bird flock attributes. It includes a unique collection of more than 6,000 high-resolution images of Demoiselle Cranes (Anthropoides virgo) feeding and nesting in the vicinity of Khichan region of Rajasthan. Particularly, each image contains an average of 190 individual birds, illustrating the complex dynamics of densely populated bird flocks on a scale that has not previously been studied. In addition, a total of 433 distinct pictures captured at Keoladeo National Park, Bharatpur provide a comprehensive representation of 34 distinct bird species belonging to various taxonomic groups. These images offer details into the diversity and the behaviour of birds in vital natural ecosystem along the migratory flyways. Additionally, we provide a set of 2,500 point-annotated samples which serve as ground truth for benchmarking various computer vision tasks like crowd counting, density estimation, segmentation, and species classification. The benchmark performance for these tasks highlight the need for tailored approaches for specific wildlife applications, which include varied conditions including views, illumination, and resolutions. With around 46.2 GBs in size encompassing data collected from two distinct nesting ground sets, it is the largest birds dataset containing detailed annotations, showcasing a substantial leap in bird research possibilities. We intend to publicly release the dataset to the research community. The database is available at: https://iab-rubric.org/resources/wildlife-dataset/birdcollect

PRP Rebooted: Advancing the State of the Art in FOND Planning

Fully Observable Non-Deterministic (FOND) planning is a variant of classical symbolic planning in which actions are nondeterministic, with an action's outcome known only upon execution. It is a popular planning paradigm with applications ranging from robot planning to dialogue-agent design and reactive synthesis. Over the last 20 years, a number of approaches to FOND planning have emerged. In this work, we establish a new state of the art, following in the footsteps of some of the most powerful FOND planners to date. Our planner, PR2, decisively outperforms the four leading FOND planners, at times by a large margin, in 17 of 18 domains that represent a comprehensive benchmark suite. Ablation studies demonstrate the impact of various techniques we introduce, with the largest improvement coming from our novel FOND-aware heuristic.

MedBench: A Large-Scale Chinese Benchmark for Evaluating Medical Large Language Models

The emergence of various medical large language models (LLMs) in the medical domain has highlighted the need for unified evaluation standards, as manual evaluation of LLMs proves to be time-consuming and labor-intensive. To address this issue, we introduce MedBench, a comprehensive benchmark for the Chinese medical domain, comprising 40,041 questions sourced from authentic examination exercises and medical reports of diverse branches of medicine. In particular, this benchmark is composed of four key components: the Chinese Medical Licensing Examination, the Resident Standardization Training Examination, the Doctor In-Charge Qualification Examination, and real-world clinic cases encompassing examinations, diagnoses, and treatments. MedBench replicates the educational progression and clinical practice experiences of doctors in Mainland China, thereby establish- ing itself as a credible benchmark for assessing the mastery of knowledge and reasoning abilities in medical language learning models. We perform extensive experiments and conduct an in-depth analysis from diverse perspectives, which culminate in the following findings: (1) Chinese medical LLMs underperform on this benchmark, highlighting the need for significant advances in clinical knowledge and diagnostic precision. (2) Several general-domain LLMs surprisingly possess considerable medical knowledge. These findings elucidate both the capabilities and limitations of LLMs within the context of MedBench, with the ultimate goal of aiding the medical research community.

DocMSU: A Comprehensive Benchmark for Document-Level Multimodal Sarcasm Understanding

Multimodal Sarcasm Understanding (MSU) has a wide range of applications in the news field such as public opinion analysis and forgery detection. However, existing MSU benchmarks and approaches usually focus on sentence-level MSU. In document-level news, sarcasm clues are sparse or small and are often concealed in long text. Moreover, compared to sentence-level comments like tweets, which mainly focus on only a few trends or hot topics (e.g., sports events), content in the news is considerably diverse. Models created for sentence-level MSU may fail to capture sarcasm clues in document-level news. To fill this gap, we present a comprehensive benchmark for Document-level Multimodal Sarcasm Understanding (DocMSU). Our dataset contains 102,588 pieces of news with text-image pairs, covering 9 diverse topics such as health, business, etc. The proposed large-scale and diverse DocMSU significantly facilitates the research of document-level MSU in real-world scenarios. To take on the new challenges posed by DocMSU, we introduce a fine-grained sarcasm comprehension method to properly align the pixel-level image features with word-level textual features in documents. Experiments demonstrate the effectiveness of our method, showing that it can serve as a baseline approach to the challenging DocMSU.

SocialStigmaQA: A Benchmark to Uncover Stigma Amplification in Generative Language Models

Current datasets for unwanted social bias auditing are limited to studying protected demographic features such as race and gender. In this work, we introduce a comprehensive benchmark that is meant to capture the amplification of social bias, via stigmas, in generative language models. Taking inspiration from social science research, we start with a documented list of 93 US-centric stigmas and curate a question-answering (QA) dataset which involves simple social situations. Our benchmark, SocialStigmaQA, contains roughly 10K prompts, with a variety of prompt styles, carefully constructed to systematically test for both social bias and model robustness. We present results for SocialStigmaQA with two open source generative language models and we find that the proportion of socially biased output ranges from 45% to 59% across a variety of decoding strategies and prompting styles. We demonstrate that the deliberate design of the templates in our benchmark (e.g., adding biasing text to the prompt or using different verbs that change the answer that indicates bias) impacts the model tendencies to generate socially biased output. Additionally, through manual evaluation, we discover problematic patterns in the generated chain-of-thought output that range from subtle bias to lack of reasoning. Warning: This paper contains examples of text which are toxic, biased, and potentially harmful.