Search

The performance differential of large language models (LLM) between languages hinders their effective deployment in many regions, inhibiting the potential economic and societal value of generative AI tools in many communities. However, the development of functional LLMs in many languages (\ie, multilingual LLMs) is bottlenecked by the lack of high-quality evaluation resources in languages other than English. Moreover, current practices in multilingual benchmark construction often translate English resources, ignoring the regional and cultural knowledge of the environments in which multilingual systems would be used. In this work, we construct an evaluation suite of 197,243 QA pairs from local exam sources to measure the capabilities of multilingual LLMs in a variety of regional contexts. Our novel resource, INCLUDE, is a comprehensive knowledge- and reasoning-centric benchmark across 44 written languages that evaluates multilingual LLMs for performance in the actual language environments where they would be deployed.

Existing Large Multimodal Models (LMMs) generally focus on only a few regions and languages. As LMMs continue to improve, it is increasingly important to ensure they understand cultural contexts, respect local sensitivities, and support low-resource languages, all while effectively integrating corresponding visual cues. In pursuit of culturally diverse global multimodal models, our proposed All Languages Matter Benchmark (ALM-bench) represents the largest and most comprehensive effort to date for evaluating LMMs across 100 languages. ALM-bench challenges existing models by testing their ability to understand and reason about culturally diverse images paired with text in various languages, including many low-resource languages traditionally underrepresented in LMM research. The benchmark offers a robust and nuanced evaluation framework featuring various question formats, including true/false, multiple choice, and open-ended questions, which are further divided into short and long-answer categories. ALM-bench design ensures a comprehensive assessment of a model's ability to handle varied levels of difficulty in visual and linguistic reasoning. To capture the rich tapestry of global cultures, ALM-bench carefully curates content from 13 distinct cultural aspects, ranging from traditions and rituals to famous personalities and celebrations. Through this, ALM-bench not only provides a rigorous testing ground for state-of-the-art open and closed-source LMMs but also highlights the importance of cultural and linguistic inclusivity, encouraging the development of models that can serve diverse global populations effectively. Our benchmark is publicly available., Comment: A Multilingual Multimodal cultural benchmark for 100 languages

Political discourse on Twitter is a moving target: politicians continuously make statements about their positions. It is therefore crucial to track their discourse on social media to understand their ideological positions and goals. However, Twitter data is also challenging to work with since it is ambiguous and often dependent on social context, and consequently, recent work on political positioning has tended to focus strongly on manifestos (parties' electoral programs) rather than social media. In this paper, we extend recently proposed methods to predict pairwise positional similarities between parties from the manifesto case to the Twitter case, using hashtags as a signal to fine-tune text representations, without the need for manual annotation. We verify the efficacy of fine-tuning and conduct a series of experiments that assess the robustness of our method for low-resource scenarios. We find that our method yields stable positioning reflective of manifesto positioning, both in scenarios with all tweets of candidates across years available and when only smaller subsets from shorter time periods are available. This indicates that it is possible to reliably analyze the relative positioning of actors forgoing manual annotation, even in the noisier context of social media., Comment: 9 pages, accepted at EMNLP (Findings) 2024

The loss of plasticity in learning agents, analogous to the solidification of neural pathways in biological brains, significantly impedes learning and adaptation in reinforcement learning due to its non-stationary nature. To address this fundamental challenge, we propose a novel approach, Neuroplastic Expansion (NE), inspired by cortical expansion in cognitive science. NE maintains learnability and adaptability throughout the entire training process by dynamically growing the network from a smaller initial size to its full dimension. Our method is designed with three key components: (1) elastic neuron generation based on potential gradients, (2) dormant neuron pruning to optimize network expressivity, and (3) neuron consolidation via experience review to strike a balance in the plasticity-stability dilemma. Extensive experiments demonstrate that NE effectively mitigates plasticity loss and outperforms state-of-the-art methods across various tasks in MuJoCo and DeepMind Control Suite environments. NE enables more adaptive learning in complex, dynamic environments, which represents a crucial step towards transitioning deep reinforcement learning from static, one-time training paradigms to more flexible, continually adapting models.

The use of deep neural networks in reinforcement learning (RL) often suffers from performance degradation as model size increases. While soft mixtures of experts (SoftMoEs) have recently shown promise in mitigating this issue for online RL, the reasons behind their effectiveness remain largely unknown. In this work we provide an in-depth analysis identifying the key factors driving this performance gain. We discover the surprising result that tokenizing the encoder output, rather than the use of multiple experts, is what is behind the efficacy of SoftMoEs. Indeed, we demonstrate that even with an appropriately scaled single expert, we are able to maintain the performance gains, largely thanks to tokenization.

Microfluidics involves the manipulation of flows at the microscale, typically requiring external power sources to generate pressure gradients. Alternatively, harnessing flows from active fluids, which are usually chaotic, has been proposed as a paradigm for the development of micro-machines. Here, by combining experimental realizations and simulations, we demonstrate that the addition of triangular-shaped inclusions into an active nematic gel can locally break the fore-aft symmetry of active turbulence and stabilize flow fields with self-pumping capabilities. The proposed strategy has enabled us to generate wall-free and self-powered microfluidic systems capable of both cargo transport and mixing along with the downstream flow. We analyze the performance of these active pumps, both isolated and within cooperative ensembles in terms of their output velocity and hydrostatic pressure buildup. Finally, we demonstrate strategies to incorporate them into specifically designed microfluidic platforms to advantageously tailor the geometry of active flows. Our results reveal new possibilities for leveraging the self-organized mechanodynamics of active fluids., Comment: Supplementary figures and videos at DOI 10.17605/OSF.IO/N8G4C

The field of active nematics has traditionally employed descriptions based on dipolar activity, with interactions that align along a single axis. However, it has been theoretically predicted that interactions with a substrate, prevalent in most biological systems, would require novel forms of activity, such as quadrupolar activity, that are governed by hydrodynamic screening. Here, by combining experiments and numerical simulations, we show that upon light-induced solidification of the underlying medium, microtubule-kinesin mixtures undergo a transformation that leads to a biphasic active suspension. Using an active lyotropic model, we prove that the transition is governed by screening effects that alter the dominant form of active stress. Specifically, the combined effect of friction and quadrupolar activity leads to a hierarchical folding that follows the intrinsic bend instability of the active nematic layer. Our results demonstrate the dynamics of the collapse of orientational order in active nematics and present a new route for controlling active matter by modifying the activity through changing the surrounding environment., Comment: 7 pages, 4 figures

Reinforcement learning from human feedback (RLHF) is fundamentally limited by the capacity of humans to correctly evaluate model output. To improve human evaluation ability and overcome that limitation this work trains "critic" models that help humans to more accurately evaluate model-written code. These critics are themselves LLMs trained with RLHF to write natural language feedback highlighting problems in code from real-world assistant tasks. On code containing naturally occurring LLM errors model-written critiques are preferred over human critiques in 63% of cases, and human evaluation finds that models catch more bugs than human contractors paid for code review. We further confirm that our fine-tuned LLM critics can successfully identify hundreds of errors in ChatGPT training data rated as "flawless", even though the majority of those tasks are non-code tasks and thus out-of-distribution for the critic model. Critics can have limitations of their own, including hallucinated bugs that could mislead humans into making mistakes they might have otherwise avoided, but human-machine teams of critics and contractors catch similar numbers of bugs to LLM critics while hallucinating less than LLMs alone.

Mixtures of Experts (MoEs) have gained prominence in (self-)supervised learning due to their enhanced inference efficiency, adaptability to distributed training, and modularity. Previous research has illustrated that MoEs can significantly boost Deep Reinforcement Learning (DRL) performance by expanding the network's parameter count while reducing dormant neurons, thereby enhancing the model's learning capacity and ability to deal with non-stationarity. In this work, we shed more light on MoEs' ability to deal with non-stationarity and investigate MoEs in DRL settings with "amplified" non-stationarity via multi-task training, providing further evidence that MoEs improve learning capacity. In contrast to previous work, our multi-task results allow us to better understand the underlying causes for the beneficial effect of MoE in DRL training, the impact of the various MoE components, and insights into how best to incorporate them in actor-critic-based DRL networks. Finally, we also confirm results from previous work.

Deep reinforcement learning (deep RL) has achieved tremendous success on various domains through a combination of algorithmic design and careful selection of hyper-parameters. Algorithmic improvements are often the result of iterative enhancements built upon prior approaches, while hyper-parameter choices are typically inherited from previous methods or fine-tuned specifically for the proposed technique. Despite their crucial impact on performance, hyper-parameter choices are frequently overshadowed by algorithmic advancements. This paper conducts an extensive empirical study focusing on the reliability of hyper-parameter selection for value-based deep reinforcement learning agents, including the introduction of a new score to quantify the consistency and reliability of various hyper-parameters. Our findings not only help establish which hyper-parameters are most critical to tune, but also help clarify which tunings remain consistent across different training regimes.

The propagation of light beams in photovoltaic pyroelectric photorefractive crystals is modelled by a specific generalization of the nonlinear Schr\"odinger equation (GNLSE). We use the variational approximation (VA) to predict the propagation of solitary-wave inputs in the crystal, finding that the VA equations involve the dilogarithm special function. The VA predicts that solitons and breathers exist, and the Vakhitov-Kolokolov criterion predicts that the solitons are stable solutions. Direct simulations of the underlying GNLSE corroborates the existence of such stable modes. The numerical solutions produce both regular breathers and ones featuring beats (long-period modulations of fast oscillations). In the latter case, the Fourier transform of amplitude oscillations reveals a nearly discrete spectrum characterizing the beats dynamics. Numerical solutions of another type demonstrate spontaneous splitting of the input pulse in two or several secondary ones., Comment: 34 pages, 28 figures, 56 references. To be published in a special issue of journal AXIOMS on the topic "Nonlinear Schroedinger equations"

Due to the widespread use of large language models (LLMs), we need to understand whether they embed a specific "worldview" and what these views reflect. Recent studies report that, prompted with political questionnaires, LLMs show left-liberal leanings (Feng et al., 2023; Motoki et al., 2024). However, it is as yet unclear whether these leanings are reliable (robust to prompt variations) and whether the leaning is consistent across policies and political leaning. We propose a series of tests which assess the reliability and consistency of LLMs' stances on political statements based on a dataset of voting-advice questionnaires collected from seven EU countries and annotated for policy issues. We study LLMs ranging in size from 7B to 70B parameters and find that their reliability increases with parameter count. Larger models show overall stronger alignment with left-leaning parties but differ among policy programs: They show a (left-wing) positive stance towards environment protection, social welfare state and liberal society but also (right-wing) law and order, with no consistent preferences in the areas of foreign policy and migration., Comment: 12 pages, TACL

Recent work has shown that deep reinforcement learning agents have difficulty in effectively using their network parameters. We leverage prior insights into the advantages of sparse training techniques and demonstrate that gradual magnitude pruning enables value-based agents to maximize parameter effectiveness. This results in networks that yield dramatic performance improvements over traditional networks, using only a small fraction of the full network parameters.

The recent rapid progress in (self) supervised learning models is in large part predicted by empirical scaling laws: a model's performance scales proportionally to its size. Analogous scaling laws remain elusive for reinforcement learning domains, however, where increasing the parameter count of a model often hurts its final performance. In this paper, we demonstrate that incorporating Mixture-of-Expert (MoE) modules, and in particular Soft MoEs (Puigcerver et al., 2023), into value-based networks results in more parameter-scalable models, evidenced by substantial performance increases across a variety of training regimes and model sizes. This work thus provides strong empirical evidence towards developing scaling laws for reinforcement learning.

Scaling analysis is a technique in computational political science that assigns a political actor (e.g. politician or party) a score on a predefined scale based on a (typically long) body of text (e.g. a parliamentary speech or an election manifesto). For example, political scientists have often used the left--right scale to systematically analyse political landscapes of different countries. NLP methods for automatic scaling analysis can find broad application provided they (i) are able to deal with long texts and (ii) work robustly across domains and languages. In this work, we implement and compare two approaches to automatic scaling analysis of political-party manifestos: label aggregation, a pipeline strategy relying on annotations of individual statements from the manifestos, and long-input-Transformer-based models, which compute scaling values directly from raw text. We carry out the analysis of the Comparative Manifestos Project dataset across 41 countries and 27 languages and find that the task can be efficiently solved by state-of-the-art models, with label aggregation producing the best results., Comment: Accepted to EMNLP 2023

Dark GRBs constitute a significant fraction of the GRB population. In this paper, we present the multiwavelength analysis of an intense two-episodic GRB 150309A observed early on to ~114 days post-burst. Despite the strong gamma-ray emission, no optical afterglow was detected for this burst. However, we discovered near-infrared afterglow ($K_{\rm S}$-band), ~5.2 hours post burst, with the CIRCE instrument mounted at the 10.4m GTC. We used Fermi observations of GRB 150309A to understand the prompt emission mechanisms and jet composition. We performed the early optical observations using the BOOTES robotic telescope and late-time afterglow observations using the GTC. A potential faint host galaxy is also detected at optical wavelength using the GTC. We modelled the potential host galaxy of GRB 150309A in order to explore the environment of the burst. The time-resolved spectral analysis of Fermi data indicates a hybrid jet composition consisting of a matter-dominated fireball and magnetic-dominated Poynting flux. GTC observations of the afterglow revealed that the counterpart of GRB 150309A was very red, with H-$K_{\rm S}$ > 2.1 mag (95 $\%$ confidence). The red counterpart was not discovered in any bluer filters of Swift UVOT, indicative of high redshift origin. This possibility was discarded based on multiple arguments, such as spectral analysis of X-ray afterglow constrain z < 4.15 and a moderate redshift value obtained using spectral energy distribution modelling of the potential galaxy. The broadband afterglow SED implies a very dusty host galaxy with deeply embedded GRB (suggesting $A_{\rm V}$ $\gtrsim$ 35 mag). The environment of GRB 150309A demands a high extinction towards the line of sight, demanding dust obscuration is the most probable origin of optical darkness and the very red afterglow of GRB 150309A. This result makes GRB 150309A the highest extinguished GRB known to date., Comment: 14 pages, 10 figures, 3 tables, accepted for publication in Astronomy & Astrophysics (A&A)

In value-based deep reinforcement learning with replay memories, the batch size parameter specifies how many transitions to sample for each gradient update. Although critical to the learning process, this value is typically not adjusted when proposing new algorithms. In this work we present a broad empirical study that suggests {\em reducing} the batch size can result in a number of significant performance gains; this is surprising, as the general tendency when training neural networks is towards larger batch sizes for improved performance. We complement our experimental findings with a set of empirical analyses towards better understanding this phenomenon., Comment: Published at NeurIPS 2023

In today's digital age, images have emerged as powerful tools for politicians to engage with their voters on social media platforms. Visual content possesses a unique emotional appeal that often leads to increased user engagement. However, research on visual communication remains relatively limited, particularly in the Global South. This study aims to bridge this gap by employing a combination of computational methods and qualitative approach to investigate the visual communication strategies employed in a dataset of 11,263 Instagram posts by 19 Brazilian presidential candidates in 2018 and 2022 national elections. Through two studies, we observed consistent patterns across these candidates on their use of visual political communication. Notably, we identify a prevalence of celebratory and positively toned images. They also exhibit a strong sense of personalization, portraying candidates connected with their voters on a more emotional level. Our research also uncovers unique contextual nuances specific to the Brazilian political landscape. We note a substantial presence of screenshots from news websites and other social media platforms. Furthermore, text-edited images with portrayals emerge as a prominent feature. In light of these results, we engage in a discussion regarding the implications for the broader field of visual political communication. This article serves as a testament to the pivotal role that Instagram has played in shaping the narrative of two fiercely polarized Brazilian elections, casting a revealing light on the ever-evolving dynamics of visual political communication in the digital age. Finally, we propose avenues for future research in the realm of visual political communication.

The landscape of information has experienced significant transformations with the rapid expansion of the internet and the emergence of online social networks. Initially, there was optimism that these platforms would encourage a culture of active participation and diverse communication. However, recent events have brought to light the negative effects of social media platforms, leading to the creation of echo chambers, where users are exposed only to content that aligns with their existing beliefs. Furthermore, malicious individuals exploit these platforms to deceive people and undermine democratic processes. To gain a deeper understanding of these phenomena, this chapter introduces a computational method designed to identify coordinated inauthentic behavior within Facebook groups. The method focuses on analyzing posts, URLs, and images, revealing that certain Facebook groups engage in orchestrated campaigns. These groups simultaneously share identical content, which may expose users to repeated encounters with false or misleading narratives, effectively forming "disinformation echo chambers." This chapter concludes by discussing the theoretical and empirical implications of these findings.

We study the quasinormal modes (QNM) for scalar, and electromagnetic perturbations in the Schwarzschild black hole with a deficit solid angle and quintessence-like matter. Using the sixth--order WKB approximation and the improved asymptotic iteration method (AIM) we can determine the dependence of the quasinormal modes on the parameters of the black hole and the parameters on the test fields. The values of the real part and imaginary parts of the quasi--normal modes increase with the decrease of the values of the deficit solid angle and density of quintessence-like matter. The quasinormal modes gotten by these two methods are in good agreement. Using the finite difference method, we obtain the time evolution profile of such perturbations in this Black Hole.

In this work, we report a direct measurement of the forces exerted by a tubulin/kinesin active nematic gel as well as its complete rheological characterization, including the quantification of its shear viscosity, {\eta}, and its activity parameter, {\alpha}. For this, we develop a novel method that allows us to rapidly photo-polymerize compliant elastic inclusions in the continuously remodelling active system. Moreover, we quantitatively settle long-standing theoretical predictions, such as a postulated relationship encoding the intrinsic time scale of the active nematic in terms of {\eta} and {\alpha}. In parallel, we infer a value for the nematic elasticity constant, K, by combining our measurements with the theorized scaling of the active length scale. On top of the microrheology capatilities, we demonstrate novel strategies for defect encapsulation, quantification of defect mechanics, and defect interactions, enabled by the versatility of the new microfabrication strategy that allows to combine elastic motifs of different shape and stiffness that are fabricated in-situ and on-time., Comment: Supplementary videos available at: https://tinyurl.com/activeUB