K. P. Subbalakshmi

Bingyang Wen, K. P. Subbalakshmi, Fan Yang

Attention mechanisms have recently demonstrated impressive performance on a range of NLP tasks, and attention scores are often used as a proxy for model explainability. However, there is a debate on whether attention weights can, in fact, be used to identify the most important inputs to a model. We approach this question from an information theoretic perspective by measuring the mutual information between the model output and the hidden states. From extensive experiments, we draw the following conclusions: (i) Additive and Deep attention mechanisms are likely to be better at preserving the information between the hidden states and the model output (compared to Scaled Dot-product); (ii) ablation studies indicate that Additive attention can actively learn to explain the importance of its input hidden representations; (iii) when attention values are nearly the same, the rank order of attention values is not consistent with the rank order of the mutual information(iv) Using Gumbel-Softmax with a temperature lower than one, tends to produce a more skewed attention score distribution compared to softmax and hence is a better choice for explainable design; (v) some building blocks are better at preserving the correlation between the ordered list of mutual information and attention weights order (for e.g., the combination of BiLSTM encoder and Additive attention). Our findings indicate that attention mechanisms do have the potential to function as a shortcut to model explanations when they are carefully combined with other model elements.

Yupeng Cao, Haohang Li, Weijin Liu et al.

Recent studies demonstrate that tool-calling capability enables large language models (LLMs) to interact with external environments for long-horizon financial tasks. While existing benchmarks have begun evaluating financial tool calling, they focus on limited scenarios and rely on call-level metrics that fail to capture trajectory-level reasoning quality. To address this gap, we introduce FinTrace, a benchmark comprising 800 expert-annotated trajectories spanning 34 real-world financial task categories across multiple difficulty levels. FinTrace employs a rubric-based evaluation protocol with nine metrics organized along four axes -- action correctness, execution efficiency, process quality, and output quality -- enabling fine-grained assessment of LLM tool-calling behavior. Our evaluation of 13 LLMs reveals that while frontier models achieve strong tool selection, all models struggle with information utilization and final answer quality, exposing a critical gap between invoking the right tools and reasoning effectively over their outputs. To move beyond diagnosis, we construct FinTrace-Training, the first trajectory-level preference dataset for financial tool-calling, containing 8,196 curated trajectories with tool-augmented contexts and preference pairs. We fine-tune Qwen-3.5-9B using supervised fine-tuning followed by direct preference optimization (DPO) and show that training on FinTrace-Training consistently improves intermediate reasoning metrics, with DPO more effectively suppressing failure modes. However, end-to-end answer quality remains a bottleneck, indicating that trajectory-level improvements do not yet fully propagate to final output quality.

Yupeng Cao, Chengyang He, Yangyang Yu et al.

Assessing the veracity of online content has become increasingly critical. Large language models (LLMs) have recently enabled substantial progress in automated veracity assessment, including automated fact-checking and claim verification systems. Typical veracity assessment pipelines break down complex claims into sub-claims, retrieve external evidence, and then apply LLM reasoning to assess veracity. However, existing methods often treat evidence retrieval as a static, isolated step and do not effectively manage or reuse retrieved evidence across claims. In this work, we propose MERMAID, a memory-enhanced multi-agent veracity assessment framework that tightly couples the retrieval and reasoning processes. MERMAID integrates agent-driven search, structured knowledge representations, and a persistent memory module within a Reason-Action style iterative process, enabling dynamic evidence acquisition and cross-claim evidence reuse. By retaining retrieved evidence in an evidence memory, the framework reduces redundant searches and improves verification efficiency and consistency. We evaluate MERMAID on three fact-checking benchmarks and two claim-verification datasets using multiple LLMs, including GPT, LLaMA, and Qwen families. Experimental results show that MERMAID achieves state-of-the-art performance while improving the search efficiency, demonstrating the effectiveness of synergizing retrieval, reasoning, and memory for reliable veracity assessment.

Yupeng Cao, Zhi Chen, Qingyun Pei et al.

In the realm of financial analytics, leveraging unstructured data, such as earnings conference calls (ECCs), to forecast stock volatility is a critical challenge that has attracted both academics and investors. While previous studies have used multimodal deep learning-based models to obtain a general view of ECCs for volatility predicting, they often fail to capture detailed, complex information. Our research introduces a novel framework: \textbf{ECC Analyzer}, which utilizes large language models (LLMs) to extract richer, more predictive content from ECCs to aid the model's prediction performance. We use the pre-trained large models to extract textual and audio features from ECCs and implement a hierarchical information extraction strategy to extract more fine-grained information. This strategy first extracts paragraph-level general information by summarizing the text and then extracts fine-grained focus sentences using Retrieval-Augmented Generation (RAG). These features are then fused through multimodal feature fusion to perform volatility prediction. Experimental results demonstrate that our model outperforms traditional analytical benchmarks, confirming the effectiveness of advanced LLM techniques in financial analysis.

Yupeng Cao, Aishwarya Muralidharan Nair, Elyon Eyimife et al.

Scientific facts are often spun in the popular press with the intent to influence public opinion and action, as was evidenced during the COVID-19 pandemic. Automatic detection of misinformation in the scientific domain is challenging because of the distinct styles of writing in these two media types and is still in its nascence. Most research on the validity of scientific reporting treats this problem as a claim verification challenge. In doing so, significant expert human effort is required to generate appropriate claims. Our solution bypasses this step and addresses a more real-world scenario where such explicit, labeled claims may not be available. The central research question of this paper is whether it is possible to use large language models (LLMs) to detect misinformation in scientific reporting. To this end, we first present a new labeled dataset SciNews, containing 2.4k scientific news stories drawn from trusted and untrustworthy sources, paired with related abstracts from the CORD-19 database. Our dataset includes both human-written and LLM-generated news articles, making it more comprehensive in terms of capturing the growing trend of using LLMs to generate popular press articles. Then, we identify dimensions of scientific validity in science news articles and explore how this can be integrated into the automated detection of scientific misinformation. We propose several baseline architectures using LLMs to automatically detect false representations of scientific findings in the popular press. For each of these architectures, we use several prompt engineering strategies including zero-shot, few-shot, and chain-of-thought prompting. We also test these architectures and prompting strategies on GPT-3.5, GPT-4, and Llama2-7B, Llama2-13B.

Zongru Shao, Xin Wang, Zhanyang Liu et al.

Recent research leverages large language models (LLMs) for early mental health detection, such as depression, often optimized with machine-generated data. However, their detection may be subject to unknown weaknesses. Meanwhile, quality control has not been applied to these generated corpora besides limited human verifications. Our goal is to systematically evaluate LLM reasoning and reveal potential weaknesses. To this end, we first provide a systematic evaluation of the reasoning over machine-generated detection and interpretation. Then we use the models' reasoning abilities to explore mitigation strategies for enhanced performance. Specifically, we do the following: A. Design an LLM instruction strategy that allows for systematic analysis of the detection by breaking down the task into several subtasks. B. Design contrastive few-shot and chain-of-thought prompts by selecting typical positive and negative examples of detection reasoning. C. Perform human annotation for the subtasks identified in the first step and evaluate the performance. D. Identify human-preferred detection with desired logical reasoning from the few-shot generation and use them to explore different optimization strategies. We conducted extensive comparisons on the DepTweet dataset across the following subtasks: 1. identifying whether the speaker is describing their own depression; 2. accurately detecting the presence of PHQ-9 symptoms, and 3. finally, detecting depression. Human verification of statistical outliers shows that LLMs demonstrate greater accuracy in analyzing and detecting explicit language of depression as opposed to implicit expressions of depression. Two optimization methods are used for performance enhancement and reduction of the statistic bias: supervised fine-tuning (SFT) and direct preference optimization (DPO). Notably, the DPO approach achieves significant performance improvement.

Yupeng Cao, Zhi Chen, Prashant Kumar et al.

The integration of Artificial Intelligence (AI) techniques, particularly large language models (LLMs), in finance has garnered increasing academic attention. Despite progress, existing studies predominantly focus on tasks like financial text summarization, question-answering, and stock movement prediction (binary classification), the application of LLMs to financial risk prediction remains underexplored. Addressing this gap, in this paper, we introduce RiskLabs, a novel framework that leverages LLMs to analyze and predict financial risks. RiskLabs uniquely integrates multimodal financial data, including textual and vocal information from Earnings Conference Calls (ECCs), market-related time series data, and contextual news data to improve financial risk prediction. Empirical results demonstrate RiskLabs' effectiveness in forecasting both market volatility and variance. Through comparative experiments, we examine the contributions of different data sources to financial risk assessment and highlight the crucial role of LLMs in this process. We also discuss the challenges associated with using LLMs for financial risk prediction and explore the potential of combining them with multimodal data for this purpose.

Mingxuan Chen, Xinqiao Chu, K. P. Subbalakshmi

The outbreak of COVID-19 has resulted in an "infodemic" that has encouraged the propagation of misinformation about COVID-19 and cure methods which, in turn, could negatively affect the adoption of recommended public health measures in the larger population. In this paper, we provide a new multimodal (consisting of images, text and temporal information) labeled dataset containing news articles and tweets on the COVID-19 vaccine. We collected 2,593 news articles from 80 publishers for one year between Feb 16th 2020 to May 8th 2021 and 24184 Twitter posts (collected between April 17th 2021 to May 8th 2021). We combine ratings from two news media ranking sites: Medias Bias Chart and Media Bias/Fact Check (MBFC) to classify the news dataset into two levels of credibility: reliable and unreliable. The combination of two filters allows for higher precision of labeling. We also propose a stance detection mechanism to annotate tweets into three levels of credibility: reliable, unreliable and inconclusive. We provide several statistics as well as other analytics like, publisher distribution, publication date distribution, topic analysis, etc. We also provide a novel architecture that classifies the news data into misinformation or truth to provide a baseline performance for this dataset. We find that the proposed architecture has an F-Score of 0.919 and accuracy of 0.882 for fake news detection. Furthermore, we provide benchmark performance for misinformation detection on tweet dataset. This new multimodal dataset can be used in research on COVID-19 vaccine, including misinformation detection, influence of fake COVID-19 vaccine information, etc.

Ning Wang, Fan Luo, Yuvraj Shivtare et al.

We propose a deep learning architecture and test three other machine learning models to automatically detect individuals that will attempt suicide within (1) 30 days and (2) six months, using their social media post data provided in the CLPsych 2021 shared task. Additionally, we create and extract three sets of handcrafted features for suicide risk detection based on the three-stage theory of suicide and prior work on emotions and the use of pronouns among persons exhibiting suicidal ideations. Extensive experimentations show that some of the traditional machine learning methods outperform the baseline with an F1 score of 0.741 and F2 score of 0.833 on subtask 1 (prediction of a suicide attempt 30 days prior). However, the proposed deep learning method outperforms the baseline with F1 score of 0.737 and F2 score of 0.843 on subtask 2 (prediction of suicide 6 months prior).

Bingyang Wen, Luis Oliveros Colon, K. P. Subbalakshmi et al.

Synthetic data generation becomes prevalent as a solution to privacy leakage and data shortage. Generative models are designed to generate a realistic synthetic dataset, which can precisely express the data distribution for the real dataset. The generative adversarial networks (GAN), which gain great success in the computer vision fields, are doubtlessly used for synthetic data generation. Though there are prior works that have demonstrated great progress, most of them learn the correlations in the data distributions rather than the true processes in which the datasets are naturally generated. Correlation is not reliable for it is a statistical technique that only tells linear dependencies and is easily affected by the dataset's bias. Causality, which encodes all underlying factors of how the real data be naturally generated, is more reliable than correlation. In this work, we propose a causal model named Causal Tabular Generative Neural Network (Causal-TGAN) to generate synthetic tabular data using the tabular data's causal information. Extensive experiments on both simulated datasets and real datasets demonstrate the better performance of our method when given the true causal graph and a comparable performance when using the estimated causal graph.

Mingxuan Chen, Ning Wang, K. P. Subbalakshmi

With social media becoming ubiquitous, information consumption from this media has also increased. However, one of the serious problems that have emerged with this increase, is the propagation of rumors. Therefore, rumor identification is a very critical task with significant implications to economy, democracy as well as public health and safety. We tackle the problem of automated detection of rumors in social media in this paper by designing a modular explainable architecture that uses both latent and handcrafted features and can be expanded to as many new classes of features as desired. This approach will allow the end user to not only determine whether the piece of information on the social media is real of a rumor, but also give explanations on why the algorithm arrived at its conclusion. Using attention mechanisms, we are able to interpret the relative importance of each of these features as well as the relative importance of the feature classes themselves. The advantage of this approach is that the architecture is expandable to more handcrafted features as they become available and also to conduct extensive testing to determine the relative influences of these features in the final decision. Extensive experimentation on popular datasets and benchmarking against eleven contemporary algorithms, show that our approach performs significantly better in terms of F-score and accuracy while also being interpretable.

Ning Wang, Mingxuan Chen, K. P. Subbalakshmi

In this work, we propose three explainable deep learning architectures to automatically detect patients with Alzheimer`s disease based on their language abilities. The architectures use: (1) only the part-of-speech features; (2) only language embedding features and (3) both of these feature classes via a unified architecture. We use self-attention mechanisms and interpretable 1-dimensional ConvolutionalNeural Network (CNN) to generate two types of explanations of the model`s action: intra-class explanation and inter-class explanation. The inter-class explanation captures the relative importance of each of the different features in that class, while the inter-class explanation captures the relative importance between the classes. Note that although we have considered two classes of features in this paper, the architecture is easily expandable to more classes because of its modularity. Extensive experimentation and comparison with several recent models show that our method outperforms these methods with an accuracy of 92.2% and F1 score of 0.952on the DementiaBank dataset while being able to generate explanations. We show by examples, how to generate these explanations using attention values.

Ning Wang, Fan Luo, Vishal Peddagangireddy et al.

Alzheimer`s disease (AD)-related global healthcare cost is estimated to be $1 trillion by 2050. Currently, there is no cure for this disease; however, clinical studies show that early diagnosis and intervention helps to extend the quality of life and inform technologies for personalized mental healthcare. Clinical research indicates that the onset and progression of Alzheimer`s disease lead to dementia and other mental health issues. As a result, the language capabilities of patient start to decline. In this paper, we show that machine learning-based unsupervised clustering of and anomaly detection with linguistic biomarkers are promising approaches for intuitive visualization and personalized early stage detection of Alzheimer`s disease. We demonstrate this approach on 10 year`s (1980 to 1989) of President Ronald Reagan`s speech data set. Key linguistic biomarkers that indicate early-stage AD are identified. Experimental results show that Reagan had early onset of Alzheimer`s sometime between 1983 and 1987. This finding is corroborated by prior work that analyzed his interviews using a statistical technique. The proposed technique also identifies the exact speeches that reflect linguistic biomarkers for early stage AD.