Recently, test-time scaling has garnered significant attention from the research community, largely due to the substantial advancements of the o1 model released by OpenAI. By allocating more computational resources during the inference phase, large language models (LLMs) can extensively explore the solution space by generating more thought tokens or diverse solutions, thereby producing more accurate responses. However, develop an o1-like reasoning approach is challenging, and researchers have been making various attempts to advance this open area of research. In this paper, we present a preliminary exploration into enhancing the reasoning abilities of LLMs through reward-guided tree search algorithms. This framework is implemented by integrating the policy model, reward model, and search algorithm. It is primarily constructed around a tree search algorithm, where the policy model navigates a dynamically expanding tree guided by a specially trained reward model. The implemented framework is denoted as STILL-1 (Slow Thinking with LLMs), marking the first model developed by our project, "Slow Thinking with LLMs". We thoroughly explore various design considerations necessary for implementing this framework and provide a detailed report of the technical aspects. To assess the effectiveness of our approach, we focus on mathematical reasoning tasks and conduct extensive evaluations on four challenging datasets, significantly enhancing the reasoning abilities of LLMs.

Modern Question Answering (QA) and Reasoning approaches based on Large Language Models (LLMs) commonly use prompting techniques, such as Chain-of-Thought (CoT), assuming the resulting generation will have a more granular exploration and reasoning over the question space and scope. However, such methods struggle with generating outputs that are faithful to the intermediate chain of reasoning produced by the model. On the other end of the spectrum, neuro-symbolic methods such as Faithful CoT (F-CoT) propose to combine LLMs with external symbolic solvers. While such approaches boast a high degree of faithfulness, they usually require a model trained for code generation and struggle with tasks that are ambiguous or hard to formalise strictly. We introduce $\textbf{F}$aithful $\textbf{L}$ogic-$\textbf{A}$ided $\textbf{R}$easoning and $\textbf{E}$xploration ($\textbf{FLARE}$), a novel interpretable approach for traversing the problem space using task decompositions. We use the LLM to plan a solution, soft-formalise the query into facts and predicates using a logic programming code and simulate that code execution using an exhaustive multi-hop search over the defined space. Our method allows us to compute the faithfulness of the reasoning process w.r.t. the generated code and analyse the steps of the multi-hop search without relying on external solvers. Our methods achieve SOTA results on $\mathbf{7}$ out of $\mathbf{9}$ diverse reasoning benchmarks. We also show that model faithfulness positively correlates with overall performance and further demonstrate that $\textbf{FLARE}$ allows pinpointing the decisive factors sufficient for and leading to the correct answer with optimal reasoning during the multi-hop search.

Feature selection is essential in the analysis of molecular systems and many other fields, but several uncertainties remain: What is the optimal number of features for a simplified, interpretable model that retains essential information? How should features with different units be aligned, and how should their relative importance be weighted? Here, we introduce the Differentiable Information Imbalance (DII), an automated method to rank information content between sets of features. Using distances in a ground truth feature space, DII identifies a low-dimensional subset of features that best preserves these relationships. Each feature is scaled by a weight, which is optimized by minimizing the DII through gradient descent. This allows simultaneously performing unit alignment and relative importance scaling, while preserving interpretability. DII can also produce sparse solutions and determine the optimal size of the reduced feature space. We demonstrate the usefulness of this approach on two benchmark molecular problems: (1) identifying collective variables that describe conformations of a biomolecule, and (2) selecting features for training a machine-learning force field. These results show the potential of DII in addressing feature selection challenges and optimizing dimensionality in various applications. The method is available in the Python library DADApy.

Recent works show that these biases may stem from uneven cultural representation in pretraining corpora. This work investigates how pretraining leads to biased culture-conditioned generations by analyzing how models associate entities with cultures based on pretraining data patterns. Additionally, the model favors generating entities with extraordinarily high frequency regardless of the conditioned culture, reflecting biases toward frequent pretraining terms irrespective of relevance. Our findings reflect trends observed specifically within OLMo-7B's pretraining data and are limited to this dataset. We make no claims about whether these results reflect real-world conditions.] In open-ended generative tasks like narrative writing or dialogue, language models often show bias against marginalized social groups based on gender, race, or culture (Gallegos et al., 2024; Manvi et al., 2024; Li et al., 2024b). Cultural bias is particularly notable due to the vast number of cultures to account for. Cultures are often unevenly represented in the pretraining corpora, with some mentioned more frequently than others, irrespective of their real-world prevalence (Li et al., 2024a). Recent studies reveal that models favor entities (Naous et al., 2023) and opinions (Ryan et al., 2024) from frequently represented cultures in pretraining while showing inadequate knowledge and templated answers for less frequent ones (Li et al., 2024b). Such biases in culture-conditioned generations can be linked to studies showing that LLMs' memorization and generalization are constrained by pretraining data imbalances. Zhang et al. (2024) find that these imbalances cause models to overgeneralize to high-frequency knowledge, overshadowing lower-frequency knowledge.

I arrive in Hermanus, a picturesque South African coastal village an hour-and-a-half from Cape Town, at about 11am on a sunny October morning. Ignoring the restaurants and art galleries on the main drag and the throngs of tourists watching southern right whales from the cliff path, I drive straight to the harbour to meet Els Vermeulen, the Belgium-born scientist who heads up the whale unit for the University of Pretoria's Mammal Research Institute. She is waiting for her colleagues to return from the last whale-tagging sortie of the 2024 season. "I would normally be out on the boat with the team," says Vermeulen, who is dressed in a bold geometric print dress and a denim jacket. "But I had to drop my kids at school and couldn't get down here early enough." The water next to the concrete pier is so clear that I can see a giant orange starfish inching its way along the rocky seabed.

Building on the foundations of language modeling in natural language processing, Next Token Prediction (NTP) has evolved into a versatile training objective for machine learning tasks across various modalities, achieving considerable success. As Large Language Models (LLMs) have advanced to unify understanding and generation tasks within the textual modality, recent research has shown that tasks from different modalities can also be effectively encapsulated within the NTP framework, transforming the multimodal information into tokens and predict the next one given the context. This survey introduces a comprehensive taxonomy that unifies both understanding and generation within multimodal learning through the lens of NTP. The proposed taxonomy covers five key aspects: Multimodal tokenization, MMNTP model architectures, unified task representation, datasets \& evaluation, and open challenges. This new taxonomy aims to aid researchers in their exploration of multimodal intelligence. An associated GitHub repository collecting the latest papers and repos is available at https://github.com/LMM101/Awesome-Multimodal-Next-Token-Prediction

The integration of advanced Natural Language Processing (NLP) methodologies and Large Language Models (LLMs) has significantly enhanced the extraction and analysis of geospatial data from multilingual texts, impacting sectors such as national and international security. This paper presents a comprehensive evaluation of leading NLP models -- SpaCy, XLM-RoBERTa, mLUKE, GeoLM -- and LLMs, specifically OpenAI's GPT 3.5 and GPT 4, within the context of multilingual geo-entity detection. Utilizing datasets from Telegram channels in English, Russian, and Arabic, we examine the performance of these models through metrics such as accuracy, precision, recall, and F1 scores, to assess their effectiveness in accurately identifying geospatial references. The analysis exposes each model's distinct advantages and challenges, underscoring the complexities involved in achieving precise geo-entity identification across varied linguistic landscapes. The conclusions drawn from this experiment aim to direct the enhancement and creation of more advanced and inclusive NLP tools, thus advancing the field of geospatial analysis and its application to global security.

Urban regeneration presents significant challenges within the context of urbanization, requiring adaptive approaches to tackle evolving needs. Leveraging advancements in large language models (LLMs), we propose Cyclical Urban Planning (CUP), a new paradigm that continuously generates, evaluates, and refines urban plans in a closed-loop. Specifically, our multi-agent LLM-based framework consists of three key components: (1) Planning, where LLM agents generate and refine urban plans based on contextual data; (2) Living, where agents simulate the behaviors and interactions of residents, modeling life in the urban environment; and (3) Judging, which involves evaluating plan effectiveness and providing iterative feedback for improvement. The cyclical process enables a dynamic and responsive planning approach. Experiments on the real-world dataset demonstrate the effectiveness of our framework as a continuous and adaptive planning process.

Less than a month before Donald Trump returns to office, two of his most ardent allies have plunged into a fierce online debate over immigration, specifically the government's visa program that allows American companies to hire so-called "highly skilled" foreign workers. The clash started on Monday with Laura Loomer, the far-right social media character known for her virulent racism, condemning Trump's decision to name Sriram Krishnan, a tech investor who was born in India, as a senior adviser on artificial intelligence. Tech leaders, including Elon Musk, weighed in to defend the practice of hiring foreign workers, specifically through the government's H-1B visa program. The debate has since devolved into a relentless string of petty insults--Loomer likened tech billionaires to "termites" at Mar-a-Lago; Musk called Loomer a troll--as well as accusations of censorship on X as retaliation. At a different point, Vivek Ramaswamy chimed in to register his support for hiring foreign workers.

Just before midnight on New Year's Eve, 25 years ago, Queen Elizabeth II stepped off a private barge to arrive at London's Millennium Dome for its grand opening ceremony. Dressed in a pumpkin-orange coat, she entered the venue with Prince Philip, taking her place alongside Tony and Cherie Blair and 12,000 guests to celebrate the dawn of a new millennium. At the stroke of midnight, Big Ben began to chime and 40 tonnes of fireworks were launched from 16 barges lined along the river. The crowd joined hands, preparing to sing Auld Lang Syne. For a few long moments, the Queen was neglected – she flapped her arms out like a toddler wanting to be lifted up, before Blair and Philip noticed her, took a hand each, and the singing began. A new century was born. One politician who wasn't in attendance at the glitzy celebration was Paddy Tipping, a Labour MP who spent the night in the Cabinet Office.