Federated Learning (FL) enables collaborative model training without sharing raw data, preserving privacy while harnessing distributed datasets. However, traditional FL systems often rely on centralized aggregating mechanisms, introducing trust issues, single points of failure, and limited mechanisms for incentivizing meaningful client contributions. These challenges are exacerbated as FL scales to train resource-intensive models, such as large language models (LLMs), requiring scalable, decentralized solutions. This paper presents a blockchain-based FL framework that addresses these limitations by integrating smart contracts and a novel hybrid incentive mechanism. The framework automates critical FL tasks, including client registration, update validation, reward distribution, and maintaining a transparent global state. The hybrid incentive mechanism combines on-chain alignment-based rewards, off-chain fairness checks, and consistency multipliers to ensure fairness, transparency, and sustained engagement. We evaluate the framework through gas cost analysis, demonstrating its feasibility for different scales of federated learning scenarios.

-- Vision-Language Models (VLMs) in visual question answering (VQA) offer a unique opportunity to enhance intra-operative decision-making, promote intuitive interactions, and significantly advancing surgical education. However, the development of VLMs for surgical VQA is challenging due to limited datasets and the risk of overfitting and catastrophic forgetting during full fine-tuning of pretrained weights. While parameter-efficient techniques like Low-Rank Adaptation (LoRA) and Matrix of Rank Adaptation (MoRA) address adaptation challenges, their uniform parameter distribution overlooks the feature hierarchy in deep networks, where earlier layers, that learn general features, require more parameters than later ones. This work introduces PitVQA++ with an open-ended PitVQA dataset and vector matrix-low-rank adaptation (V ector-MoLoRA), an innovative VLM fine-tuning approach for adapting GPT -2 to pituitary surgery. Open-Ended PitVQA comprises around 101,803 frames from 25 procedural videos with 745,972 question-answer sentence pairs, covering key surgical elements such as phase and step recognition, context understanding, tool detection, localization, and interactions recognition. V ector-MoLoRA incorporates the principles of LoRA and MoRA to develop a matrix-low-rank adaptation strategy that employs vector ranking to allocate more parameters to earlier layers, gradually reducing them in the later layers. Furthermore, our risk-coverage analysis highlights its enhanced reliability and trustworthiness in handling uncertain predictions.

This paper describes our submissions to the TREC 2024 PLABA track with the aim to simplify biomedical abstracts for a K8 - level audience (13 - 14 years old students). We tested three approaches using OpenAI's gpt - 4o and gpt - 4o - mini models: baseline prompt engineering, a two - AI agent approach, and fine - tuning. Adaptations were evaluated using qualitative metrics ( 5 - point Likert scales for simplicity, accuracy, completeness, and brevity) and quantitative readability scores (Flesch - Kincaid grade level, SMOG Index). Results indicate d that the two - agent approach and baseline prompt engineering with gpt - 4o - mini models show superior qualitative performance, while fine - tuned models excelled in accuracy and completeness but were less simple. The evaluation results demonstrated that prompt engineering with gpt - 4o - mini outperforms iterative improvement strategies via two - agent approach as well as fine - tuning with gpt - 4o. We intend to expand our investigation of the results and explore advanced evaluations.

Modern text classification methods heavily rely on contextual embeddings from large language models (LLMs). Compared to human-engineered features, these embeddings provide automatic and effective representations for classification model training. However, they also introduce a challenge: we lose the ability to manually remove unintended features, such as sensitive or task-irrelevant features, to guarantee regulatory compliance or improve the generalizability of classification models. This limitation arises because LLM embeddings are opaque and difficult to interpret. In this paper, we propose a novel framework to identify and regularize unintended features in the LLM latent space. Specifically, we first pre-train a sparse autoencoder (SAE) to extract interpretable features from LLM latent spaces. To ensure the SAE can capture task-specific features, we further fine-tune it on task-specific datasets. In training the classification model, we propose a simple and effective regularizer, by minimizing the similarity between the classifier weights and the identified unintended feature, to remove the impacts of these unintended features toward classification. We evaluate the proposed framework on three real-world tasks, including toxic chat detection, reward modeling, and disease diagnosis. Results show that the proposed framework can significantly improve the classifier's generalizability by regularizing those features that are not semantically correlated to each task. This work pioneers controllable text classification on LLM latent spaces by leveraging interpreted features to address generalizability, fairness, and privacy challenges. We will release our code and data once accepted.

Two commonly-employed strategies to combat the rise of misinformation on social media are (i) fact-checking by professional organisations and (ii) community moderation by platform users. Policy changes by Twitter/X and, more recently, Meta, signal a shift away from partnerships with fact-checking organisations and towards an increased reliance on crowdsourced community notes. However, the extent and nature of dependencies between fact-checking and helpful community notes remain unclear. To address these questions, we use language models to annotate a large corpus of Twitter/X community notes with attributes such as topic, cited sources, and whether they refute claims tied to broader misinformation narratives. Our analysis reveals that community notes cite fact-checking sources up to five times more than previously reported. Fact-checking is especially crucial for notes on posts linked to broader narratives, which are twice as likely to reference fact-checking sources compared to other sources. In conclusion, our results show that successful community moderation heavily relies on professional fact-checking.

Multiple choice question answering (MCQA) is popular for LLM evaluation due to its simplicity and human-like testing, but we argue for its reform. We first reveal flaws in MCQA's format, as it struggles to: 1) test generation/subjectivity; 2) match LLM use cases; and 3) fully test knowledge. We instead advocate for generative formats based on human testing-where LLMs construct and explain answers-better capturing user needs and knowledge while remaining easy to score. We then show even when MCQA is a useful format, its datasets suffer from: leakage; unanswerability; shortcuts; and saturation. In each issue, we give fixes from education, like rubrics to guide MCQ writing; scoring methods to bridle guessing; and Item Response Theory to build harder MCQs. Lastly, we discuss LLM errors in MCQA-robustness, biases, and unfaithful explanations-showing how our prior solutions better measure or address these issues. While we do not need to desert MCQA, we encourage more efforts in refining the task based on educational testing, advancing evaluations.

Large Language Models (LLMs) have achieved significant advances in natural language processing, yet their potential for high-stake political decision-making remains largely unexplored. This paper addresses the gap by focusing on the application of LLMs to the United Nations (UN) decision-making process, where the stakes are particularly high and political decisions can have far-reaching consequences. We introduce a novel dataset comprising publicly available UN Security Council (UNSC) records from 1994 to 2024, including draft resolutions, voting records, and diplomatic speeches. Using this dataset, we propose the United Nations Benchmark (UNBench), the first comprehensive benchmark designed to evaluate LLMs across four interconnected political science tasks: co-penholder judgment, representative voting simulation, draft adoption prediction, and representative statement generation. These tasks span the three stages of the UN decision-making process--drafting, voting, and discussing--and aim to assess LLMs' ability to understand and simulate political dynamics. Our experimental analysis demonstrates the potential and challenges of applying LLMs in this domain, providing insights into their strengths and limitations in political science. This work contributes to the growing intersection of AI and political science, opening new avenues for research and practical applications in global governance. The UNBench Repository can be accessed at: https://github.com/yueqingliang1/UNBench.

We present a hierarchy of natural language understanding abilities and argue for the importance of moving beyond assessments of understanding at the lexical and sentence levels to the discourse level. We propose the task of anaphora accessibility as a diagnostic for assessing discourse understanding, and to this end, present an evaluation dataset inspired by theoretical research in dynamic semantics. We evaluate human and LLM performance on our dataset and find that LLMs and humans align on some tasks and diverge on others. Such divergence can be explained by LLMs' reliance on specific lexical items during language comprehension, in contrast to human sensitivity to structural abstractions.

Recent advances in vision language models (VLM) have been driven by contrastive models such as CLIP, which learn to associate visual information with their corresponding text descriptions. However, these models have limitations in understanding complex compositional scenes involving multiple objects and their spatial relationships. To address these challenges, we propose a novel approach that diverges from commonly used strategies, which rely on the design of hard-negative augmentations. Instead, our work focuses on integrating inductive biases into pre-trained CLIP-like models to improve their compositional understanding without using any additional hard-negatives. To that end, we introduce a binding module that connects a scene graph, derived from a text description, with a slot-structured image representation, facilitating a structured similarity assessment between the two modalities. We also leverage relationships as text-conditioned visual constraints, thereby capturing the intricate interactions between objects and their contextual relationships more effectively. Our resulting model not only enhances the performance of CLIP-based models in multi-object compositional understanding but also paves the way towards more accurate and sample-efficient image-text matching of complex scenes.

A key aspect of alignment is the proper use of within-document evidence to construct document-level decisions. We analyze the relationship between the retrieval and interpretation of within-document evidence for large language models in a few-shot setting. Specifically, we measure the extent to which model prediction errors are associated with evidence retrieval errors with respect to gold-standard human-annotated extractive evidence for five datasets, using two popular closed proprietary models. We perform two ablation studies to investigate when both label prediction and evidence retrieval errors can be attributed to qualities of the relevant evidence. We find that there is a strong empirical relationship between model prediction and evidence retrieval error, but that evidence retrieval error is mostly not associated with evidence interpretation error-a hopeful sign for downstream applications built on this mechanism.