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.

The ability of large language models (LLMs) to follow instructions is crucial for their practical applications, yet the underlying mechanisms remain poorly understood. This paper presents a novel framework that leverages sparse autoencoders (SAE) to interpret how instruction following works in these models. We demonstrate how the features we identify can effectively steer model outputs to align with given instructions. Through analysis of SAE latent activations, we identify specific latents responsible for instruction following behavior. Our findings reveal that instruction following capabilities are encoded by a distinct set of instruction-relevant SAE latents. These latents both show semantic proximity to relevant instructions and demonstrate causal effects on model behavior. Our research highlights several crucial factors for achieving effective steering performance: precise feature identification, the role of final layer, and optimal instruction positioning. Additionally, we demonstrate that our methodology scales effectively across SAEs and LLMs of varying sizes.

Reasoning abilities of LLMs have been a key focus in recent years. One challenging reasoning domain with interesting nuances is legal reasoning, which requires careful application of rules, and precedents while balancing deductive and analogical reasoning, and conflicts between rules. Although there have been a few works on using LLMs for legal reasoning, their focus has been on overall accuracy. In this paper, we dig deeper to do a step-by-step analysis and figure out where they commit errors. We use the college-level Multiple Choice Question-Answering (MCQA) task from the \textit{Civil Procedure} dataset and propose a new error taxonomy derived from initial manual analysis of reasoning chains with respect to several LLMs, including two objective measures: soundness and correctness scores. We then develop an LLM-based automated evaluation framework to identify reasoning errors and evaluate the performance of LLMs. The computation of soundness and correctness on the dataset using the auto-evaluator framework reveals several interesting insights. Furthermore, we show that incorporating the error taxonomy as feedback in popular prompting techniques marginally increases LLM performance. Our work will also serve as an evaluation framework that can be used in detailed error analysis of reasoning chains for logic-intensive complex tasks.

Enhancing the reasoning capabilities of large language models (LLMs) remains a key challenge, especially for tasks that require complex, multi-step decision-making. Humans excel at these tasks by leveraging deliberate planning with an internal world model to simulate the potential outcomes of various actions. Inspired by this, we propose a novel multi-step reasoning framework for LLMs, referred to as Structure-aware Planning with Accurate World Model (SWAP). Unlike previous approaches that rely solely on Chain-of-Thought (CoT) reasoning in natural language, SWAP incorporates structural information to guide the reasoning process via a world model and provides a soft verification mechanism over the steps. Moreover, SWAP overcomes the challenge of accurate world state predictions in complex reasoning tasks by introducing a Generator-Discriminator architecture, which enables more reliable world modeling. Specifically, the generator predicts the next state, and the discriminator ensures alignment with the logical consistency required by the problem context. SWAP also encourages the policy model to explore a broad range of potential actions to prevent premature convergence. By resolving the bottlenecks of generation diversity for both actions and states using diversity-based modeling (DBM) and improving discrimination accuracy through contrastive ranking (CR), SWAP significantly enhances the reasoning performance of LLMs. We evaluate SWAP across diverse reasoning-intensive benchmarks including math reasoning, logical reasoning, and coding tasks. Extensive experiments demonstrate that SWAP achieves substantial improvements over the baselines and consistently outperforms existing methods.

Modern deep learning models often make predictions by focusing on irrelevant areas, leading to biased performance and limited generalization. Existing methods aimed at rectifying model attention require explicit labels for irrelevant areas or complex pixel-wise ground truth attention maps. We present CRAYON (Correcting Reasoning with Annotations of Yes Or No), offering effective, scalable, and practical solutions to rectify model attention using simple yes-no annotations. CRAYON empowers classical and modern model interpretation techniques to identify and guide model reasoning: CRAYON-ATTENTION directs classic interpretations based on saliency maps to focus on relevant image regions, while CRAYON-PRUNING removes irrelevant neurons identified by modern concept-based methods to mitigate their influence. Through extensive experiments with both quantitative and human evaluation, we showcase CRAYON's effectiveness, scalability, and practicality in refining model attention. CRAYON achieves state-of-the-art performance, outperforming 12 methods across 3 benchmark datasets, surpassing approaches that require more complex annotations.

Large Language Models (LLMs) suffer from hallucinations, referring to the nonfactual information in generated content, despite their superior capacities across tasks. Meanwhile, knowledge editing has been developed as a new popular paradigm to correct the erroneous factual knowledge encoded in LLMs with the advantage of avoiding retraining from scratch. However, one common issue of existing evaluation datasets for knowledge editing is that they do not ensure LLMs actually generate hallucinated answers to the evaluation questions before editing. When LLMs are evaluated on such datasets after being edited by different techniques, it is hard to directly adopt the performance to assess the effectiveness of different knowledge editing methods in correcting hallucinations. Thus, the fundamental question remains insufficiently validated: Can knowledge editing really correct hallucinations in LLMs? We proposed HalluEditBench to holistically benchmark knowledge editing methods in correcting real-world hallucinations. First, we rigorously construct a massive hallucination dataset with 9 domains, 26 topics and more than 6, 000 hallucinations. Then, we assess the performance of knowledge editing methods in a holistic way on five dimensions including Efficacy, Generalization, Portability, Locality, and Robustness. Through HalluEditBench, we have provided new insights into the potentials and limitations of different knowledge editing methods in correcting hallucinations, which could inspire future improvements and facilitate the progress in the field of knowledge editing. Considering Table 1: Performance measured by Accuracy (%) the high cost of retraining LLMs from scratch, of Llama2-7B before editing ("Pre-edit") and after knowledge editing has been designed as a new applying typical knowledge editing methods ("Postedit") paradigm to correct erroneous or outdated factual on common existing evaluation datasets. When such datasets are adopted to evaluate the performance of LLMs after being edited, it is hard to directly use the scores to judge the effectiveness of different knowledge editing techniques in correcting hallucinations, which is the motivation of applying knowledge editing to LLMs. To better illustrate this point, following the evaluation setting in (Zhang et al., 2024e), we conducted a preliminary study to examine the pre-edit and post-edit performances of Llama2-7B on the aforementioned Who is the Chief Scientist of OpenAI? Who is the Chief Scientist of OpenAI? Who is the Chief Scientist of OpenAI?

Performative prediction (PP) is a framework that captures distribution shifts that occur during the training of machine learning models due to their deployment. As the trained model is used, its generated data could cause the model to evolve, leading to deviations from the original data distribution. The impact of such model-induced distribution shifts in the federated learning (FL) setup remains unexplored despite being increasingly likely to transpire in real-life use cases. Although Jin et al. (2024) recently extended PP to FL in a straightforward manner, the resulting model only converges to a performative stable point, which may be far from optimal. The methods in Izzo et al. (2021); Miller et al. (2021) can find a performative optimal point in centralized settings, but they require the performative risk to be convex and the training data to be noiseless, assumptions often violated in realistic FL systems. This paper overcomes all of these shortcomings and proposes Performative robust optimal Federated Learning (ProFL), an algorithm that finds performative optimal points in FL from noisy and contaminated data. We present the convergence analysis under the Polyak-Lojasiewicz condition, which applies to non-convex objectives. Extensive experiments on multiple datasets validate our proposed algorithms' efficiency.

Probing learned concepts in large language models (LLMs) is crucial for understanding how semantic knowledge is encoded internally. Training linear classifiers on probing tasks is a principle approach to denote the vector of a certain concept in the representation space. However, the single vector identified for a concept varies with both data and training, making it less robust and weakening its effectiveness in real-world applications. To address this challenge, we propose an approach to approximate the subspace representing a specific concept. Built on linear probing classifiers, we extend the concept vectors into Gaussian Concept Subspace (GCS). We demonstrate GCS's effectiveness through measuring its faithfulness and plausibility across multiple LLMs with different sizes and architectures. Additionally, we use representation intervention tasks to showcase its efficacy in real-world applications such as emotion steering. Experimental results indicate that GCS concept vectors have the potential to balance steering performance and maintaining the fluency in natural language generation tasks.

Modern machine learning models trained on losses based on point predictions are prone to be overconfident in their predictions [Guo et al., 2017]. The Conformal Prediction (CP) framework [Vovk et al., 2005] provides a mechanism for generating statistically sound post hoc prediction sets (or intervals, in case of continuous outcomes) with coverage guarantees under mild assumptions. The usual assumption made in CP is that data are exchangeable, i.e., the joint distribution of the data is invariant to permutations of the data points. CP's guarantees are distribution-free and can be added post hoc to arbitrary black-box predictor scores, making them ideal candidates for quantifying uncertainty in complex models, such as neural networks. Network-structured data such as social networks, transportation networks, and biological networks are ubiquitous in modern data science applications.

The development of Large Language Models (LLMs) relies on extensive text corpora, which are often unevenly distributed across languages. This imbalance results in LLMs performing significantly better on high-resource languages like English, German, and French, while their capabilities in low-resource languages remain inadequate. Currently, there is a lack of quantitative methods to evaluate the performance of LLMs in these low-resource languages. To address this gap, we propose the Language Ranker, an intrinsic metric designed to benchmark and rank languages based on LLM performance using internal representations. By comparing the LLM's internal representation of various languages against a baseline derived from English, we can assess the model's multilingual capabilities in a robust and language-agnostic manner. Our analysis reveals that high-resource languages exhibit higher similarity scores with English, demonstrating superior performance, while low-resource languages show lower similarity scores, underscoring the effectiveness of our metric in assessing language-specific capabilities. Besides, the experiments show that there is a strong correlation between the LLM's performance in different languages and the proportion of those languages in its pre-training corpus. These insights underscore the efficacy of the Language Ranker as a tool for evaluating LLM performance across different languages, particularly those with limited resources.