As large language models (LLM) become more and more capable in languages other than English, it is important to collect benchmark datasets in order to evaluate their multilingual performance, including on tasks like machine translation (MT). In this work, we extend the WMT24 dataset to cover 55 languages by collecting new human-written references and post-edits for 46 new languages and dialects in addition to post-edits of the references in 8 out of 9 languages in the original WMT24 dataset. The dataset covers four domains: literary, news, social, and speech. We benchmark a variety of MT providers and LLMs on the collected dataset using automatic metrics and find that LLMs are the best-performing MT systems in all 55 languages. These results should be confirmed using a human-based evaluation, which we leave for future work.

Previous studies primarily utilize MLP neurons as units of analysis for understanding the mechanisms of factual knowledge in Language Models (LMs); however, neurons suffer from polysemanticity, leading to limited knowledge expression and poor interpretability. In this paper, we first conduct preliminary experiments to validate that Sparse Autoencoders (SAE) can effectively decompose neurons into features, which serve as alternative analytical units. With this established, our core findings reveal three key advantages of features over neurons: (1) Features exhibit stronger influence on knowledge expression and superior interpretability. (2) Features demonstrate enhanced monosemanticity, showing distinct activation patterns between related and unrelated facts. (3) Features achieve better privacy protection than neurons, demonstrated through our proposed FeatureEdit method, which significantly outperforms existing neuron-based approaches in erasing privacy-sensitive information from LMs.Code and dataset will be available.

Advances in machine learning and the increasing availability of high-dimensional data have led to the proliferation of social science research that uses the predictions of machine learning models as proxies for measures of human activity or environmental outcomes. However, prediction errors from machine learning models can lead to bias in the estimates of regression coefficients. In this paper, we show how this bias can arise, propose a test for detecting bias, and demonstrate the use of an adversarial machine learning algorithm in order to de-bias predictions. These methods are applicable to any setting where machine-learned predictions are the dependent variable in a regression. We conduct simulations and empirical exercises using ground truth and satellite data on forest cover in Africa. Using the predictions from a naive machine learning model leads to biased parameter estimates, while the predictions from the adversarial model recover the true coefficients.

Four students killed in a car crash were honoured at a university as basketball matches resumed for the first time since the incident. Makyle Bayley, 22, Eva Darold-Tchikaya, 21, Anthony "TJ" Hibbert, 24 and Daljang Wol, 22, died when a car crashed into a building on Magdalen Street, Colchester on 1 February. Mr Hibbert and Mr Wol played for the Essex Rebels, who dedicated Saturday's fixtures to the victims and held an applause in their memory. University of Essex director of sport Dave Parry said: "We've lost four really loved members of our university and sporting community, who gave so much to their friends and others." Mr Bayley was a member of the British Universities and Colleges Sport (BUCS) basketball team, while Ms Darold-Tchikaya was a member of the Essex Blades dance club and other societies.Dawid Wojtowicz/BBCSaturday's basketball fixtures at the University of Essex were dedicated to the victimsDawid Wojtowicz/BBCIt was the first time matches had been played there since the incident Last week, more than 1,000 people including students, staff and relatives of the victims attended a gathering.

Solving complex tasks in a single attempt is challenging for large language models (LLMs). Iterative interaction with the environment and feedback is often required to achieve success, making effective feedback utilization a critical topic. Existing approaches either struggle with length generalization or rely on naive retries without leveraging prior information. In this paper, we introduce FTTT, a novel paradigm that formulates feedback utilization as an optimization problem at test time. Additionally, we propose a learnable test-time optimizer, OpTune, to effectively exploit feedback. Experiments on two LLMs across four reasoning datasets demonstrate that FTTT and OpTune achieve superior scalability and performance.

Visual-Language Models (VLMs) have shown remarkable performance across various tasks, particularly in recognizing geographic information from images. However, significant challenges remain, including biases and privacy concerns. To systematically address these issues in the context of geographic information recognition, we introduce a benchmark dataset consisting of 1,200 images paired with detailed geographic metadata. Evaluating four VLMs, we find that while these models demonstrate the ability to recognize geographic information from images, achieving up to $53.8\%$ accuracy in city prediction, they exhibit significant regional biases. Specifically, performance is substantially higher for economically developed and densely populated regions compared to less developed ($-12.5\%$) and sparsely populated ($-17.0\%$) areas. Moreover, the models exhibit regional biases, frequently overpredicting certain locations; for instance, they consistently predict Sydney for images taken in Australia. The strong performance of VLMs also raises privacy concerns, particularly for users who share images online without the intent of being identified. Our code and dataset are publicly available at https://github.com/uscnlp-lime/FairLocator.

Retrieval-augmented generation (RAG) enhances large language models (LLMs) for domain-specific question-answering (QA) tasks by leveraging external knowledge sources. However, traditional RAG systems primarily focus on relevance-based retrieval and often struggle with redundancy, especially when reasoning requires connecting information from multiple sources. This paper introduces Vendi-RAG, a framework based on an iterative process that jointly optimizes retrieval diversity and answer quality. This joint optimization leads to significantly higher accuracy for multi-hop QA tasks. Vendi-RAG leverages the Vendi Score (VS), a flexible similarity-based diversity metric, to promote semantic diversity in document retrieval. It then uses an LLM judge that evaluates candidate answers, generated after a reasoning step, and outputs a score that the retriever uses to balance relevance and diversity among the retrieved documents during each iteration. Experiments on three challenging datasets -- HotpotQA, MuSiQue, and 2WikiMultiHopQA -- demonstrate Vendi-RAG's effectiveness in multi-hop reasoning tasks. The framework achieves significant accuracy improvements over traditional single-step and multi-step RAG approaches, with accuracy increases reaching up to +4.2% on HotpotQA, +4.1% on 2WikiMultiHopQA, and +1.3% on MuSiQue compared to Adaptive-RAG, the current best baseline. The benefits of Vendi-RAG are even more pronounced as the number of retrieved documents increases. Finally, we evaluated Vendi-RAG across different LLM backbones, including GPT-3.5, GPT-4, and GPT-4o-mini, and observed consistent improvements, demonstrating that the framework's advantages are model-agnostic.

Transformer-based embedding methods have dominated the field of sentence representation in recent years. Although they have achieved remarkable performance on NLP missions, such as semantic textual similarity (STS) tasks, their black-box nature and large-data-driven training style have raised concerns, including issues related to bias, trust, and safety. Many efforts have been made to improve the interpretability of embedding models, but these problems have not been fundamentally resolved. To achieve inherent interpretability, we propose a purely white-box and human-like sentence representation network, PropNet. Inspired by findings from cognitive science, PropNet constructs a hierarchical network based on the propositions contained in a sentence. While experiments indicate that PropNet has a significant gap compared to state-of-the-art (SOTA) embedding models in STS tasks, case studies reveal substantial room for improvement. Additionally, PropNet enables us to analyze and understand the human cognitive processes underlying STS benchmarks.

Deep Reinforcement Learning (RL) is remarkably effective in addressing sequential resource allocation problems in domains such as healthcare, public policy, and resource management. However, deep RL policies often lack transparency and adaptability, challenging their deployment alongside human decision-makers. In contrast, Language Agents, powered by large language models (LLMs), provide human-understandable reasoning but may struggle with effective decision making. To bridge this gap, we propose Rule-Bottleneck Reinforcement Learning (RBRL), a novel framework that jointly optimizes decision and explanations. At each step, RBRL generates candidate rules with an LLM, selects among them using an attention-based RL policy, and determines the environment action with an explanation via chain-of-thought reasoning. The RL rule selection is optimized using the environment rewards and an explainability metric judged by the LLM. Evaluations in real-world scenarios highlight RBRL's competitive performance with deep RL and efficiency gains over LLM fine-tuning. A survey further confirms the enhanced quality of its explanations.

Mass spectrometry, especially so-called tandem mass spectrometry, is commonly used to assess the chemical diversity of samples. The resulting mass fragmentation spectra are representations of molecules of which the structure may have not been determined. This poses the challenge of experimentally determining or computationally predicting molecular structures from mass spectra. An alternative option is to predict molecular properties or molecular similarity directly from spectra. Various methodologies have been proposed to embed mass spectra for further use in machine learning tasks. However, these methodologies require preprocessing of the spectra, which often includes binning or sub-sampling peaks with the main reasoning of creating uniform vector sizes and removing noise. Here, we investigate two alternatives to the binning of mass spectra before down-stream machine learning tasks, namely, set-based and graph-based representations. Comparing the two proposed representations to train a set transformer and a graph neural network on a regression task, respectively, we show that they both perform substantially better than a multilayer perceptron trained on binned data.