Large Language Models (LLMs) are rapidly being adopted by users across the globe, who interact with them in a diverse range of languages. At the same time, there are well-documented imbalances in the training data and optimisation objectives of this technology, raising doubts as to whether LLMs can represent the cultural diversity of their broad user base. In this study, we look at LLMs and cultural values and examine how prompt language and cultural framing influence model responses and their alignment with human values in different countries. We probe 10 LLMs with 63 items from the Hofstede Values Survey Module and World Values Survey, translated into 11 languages, and formulated as prompts with and without different explicit cultural perspectives. Our study confirms that both prompt language and cultural perspective produce variation in LLM outputs, but with an important caveat: While targeted prompting can, to a certain extent, steer LLM responses in the direction of the predominant values of the corresponding countries, it does not overcome the models' systematic bias toward the values associated with a restricted set of countries in our dataset: the Netherlands, Germany, the US, and Japan. All tested models, regardless of their origin, exhibit remarkably similar patterns: They produce fairly neutral responses on most topics, with selective progressive stances on issues such as social tolerance. Alignment with cultural values of human respondents is improved more with an explicit cultural perspective than with a targeted prompt language. Unexpectedly, combining both approaches is no more effective than cultural framing with an English prompt. These findings reveal that LLMs occupy an uncomfortable middle ground: They are responsive enough to changes in prompts to produce variation, but too firmly anchored to specific cultural defaults to adequately represent cultural diversity.

In this study, we experiment with the ability of LLMs to do temporal reasoning. Using a Norwegian book from 1940 containing trivia questions, we prompt the LLMs to answer the questions as if it were 1940. We also pose the questions in both English and Norwegian. Correct answers are often presented as sentences, and grading is done by means of LLM-as-judge, with sampled checks by a native speaker. Prompting in English consistently gave better results than in Norwegian, an unexpected result. In contrast, using larger LLMs improved results. We tested the DeepSeek-R1, Gemma3, Qwen3, and Llama3.1 model families, and also the largest available LLM especially crafted for Norwegian.

Recent advances in Large Language Models (LLMs) have highlighted the critical importance of extending context length, yet the quadratic complexity of attention mechanisms poses significant challenges for efficient long-context modeling. KV cache compression has emerged as a key approach to address this challenge. Through extensive empirical analysis, we reveal a fundamental yet previously overlooked asymmetry in KV caches: while adjacent keys receive similar attention weights ({\it local homogeneity}), adjacent values demonstrate distinct {\it heterogeneous} distributions. This key-value asymmetry reveals a critical limitation in existing compression methods that treat keys and values uniformly. To address the limitation, we propose a training-free compression framework (AsymKV) that combines homogeneity-based key merging with a mathematically proven lossless value compression. Extensive experiments demonstrate that AsymKV consistently outperforms existing long-context methods across various tasks and base models. For example, on LLaMA3.1-8B, AsymKV achieves an average score of 43.95 on LongBench, surpassing SOTA methods like H$_2$O (38.89) by a large margin.Our code can be found in this link:https://github.com/the-scale-lab/Asymkv.

Recent work suggests that large language models (LLMs) can perform multi-hop reasoning implicitly -- producing correct answers without explicitly verbalizing intermediate steps -- but the underlying mechanisms remain poorly understood. In this paper, we study how such implicit reasoning emerges by training transformers from scratch in a controlled symbolic environment. Our analysis reveals a three-stage developmental trajectory: early memorization, followed by in-distribution generalization, and eventually cross-distribution generalization. We find that training with atomic triples is not necessary but accelerates learning, and that second-hop generalization relies on query-level exposure to specific compositional structures. To interpret these behaviors, we introduce two diagnostic tools: cross-query semantic patching, which identifies semantically reusable intermediate representations, and a cosine-based representational lens, which reveals that successful reasoning correlates with the cosine-base clustering in hidden space. This clustering phenomenon in turn provides a coherent explanation for the behavioral dynamics observed across training, linking representational structure to reasoning capability. These findings provide new insights into the interpretability of implicit multi-hop reasoning in LLMs, helping to clarify how complex reasoning processes unfold internally and offering pathways to enhance the transparency of such models.

Abstract--T okenization is fundamental in assembly code analysis, impacting intrinsic characteristics like vocabulary size, semantic coverage, and extrinsic performance in downstream tasks. Despite its significance, tokenization in the context of assembly code remains an underexplored area. This study aims to address this gap by evaluating the intrinsic properties of Natural Language Processing (NLP) tokenization models and parameter choices, such as vocabulary size. We explore prepro-cessing customization options and pre-tokenization rules tailored to the unique characteristics of assembly code. Additionally, we assess their impact on downstream tasks like function signature prediction--a critical problem in binary code analysis. T o this end, we conduct a thorough study on various tokeniza-tion models, systematically analyzing their efficiency in encoding assembly instructions and capturing semantic nuances. Through intrinsic evaluations, we compare tokenizers based on tokeniza-tion efficiency, vocabulary compression, and representational fidelity for assembly code. Using state-of-the-art pre-trained models such as the decoder-only Large Language Model (LLM) Llama 3.2, the encoder-only transformer BERT, and the encoder-decoder model BART, we evaluate the effectiveness of these tokenizers across multiple performance metrics. Preliminary findings indicate that tokenizer choice significantly influences downstream performance, with intrinsic metrics providing partial but incomplete predictability of extrinsic evaluation outcomes. These results reveal complex trade-offs between intrinsic tokenizer properties and their utility in practical assembly code tasks. Ultimately, this study provides valuable insights into optimizing tokenization models for low-level code analysis, contributing to the robustness and scalability of Natural Language Model (NLM)-based binary analysis workflows. Tokenization is critical in transforming raw input data into structured representations, a process of utmost importance for Machine Learning (ML) and NLM model tasks [1]-[3]. While tokenization strategies have been studied extensively for natural [4] and high-level programming languages [5], assembly code presents unique challenges due to its low-level operations, diverse instruction sets, and non-standardized syntax across architectures. These challenges highlight the need for specialized tokenization techniques that effectively capture assembly code's structural and semantic intricacies [2]. Despite its importance, the role of tokenization in assembly code processing remains underexplored, particularly in its impact on downstream tasks involving modern NLMs. Recent research underscores the significant influence of tokenization on NLM model performance.

Large Language Models (LLMs) play a central role in modern artificial intelligence, yet their development has been primarily focused on English, resulting in limited support for other languages. We present PLLuM (Polish Large Language Model), the largest open-source family of foundation models tailored specifically for the Polish language. Developed by a consortium of major Polish research institutions, PLLuM addresses the need for high-quality, transparent, and culturally relevant language models beyond the English-centric commercial landscape. We describe the development process, including the construction of a new 140-billion-token Polish text corpus for pre-training, a 77k custom instructions dataset, and a 100k preference optimization dataset. A key component is a Responsible AI framework that incorporates strict data governance and a hybrid module for output correction and safety filtering. We detail the models' architecture, training procedures, and alignment techniques for both base and instruction-tuned variants, and demonstrate their utility in a downstream task within public administration. By releasing these models publicly, PLLuM aims to foster open research and strengthen sovereign AI technologies in Poland.

Reasoning language models perform well on complex tasks but are costly to deploy due to their size and long reasoning traces. We propose a routing approach that assigns each problem to the smallest model likely to solve it, reducing compute without sacrificing accuracy. Using intermediate representations from s1.1-32B, we train lightweight predictors of problem difficulty or model correctness to guide routing across a pool of reasoning models. On diverse math benchmarks, routing improves efficiency over random assignment and matches s1.1-32B's performance while using significantly less compute. Our results demonstrate that difficulty-aware routing is effective for cost-efficient deployment of reasoning models.

Large Language Models (LLMs) show great promise as a powerful tool for scientific literature exploration. However, their effectiveness in providing scientifically accurate and comprehensive answers to complex questions within specialized domains remains an active area of research. Using the field of high-temperature cuprates as an exemplar, we evaluate the ability of LLM systems to understand the literature at the level of an expert. We construct an expert-curated database of 1,726 scientific papers that covers the history of the field, and a set of 67 expert-formulated questions that probe deep understanding of the literature. We then evaluate six different LLM-based systems for answering these questions, including both commercially available closed models and a custom retrieval-augmented generation (RAG) system capable of retrieving images alongside text. Experts then evaluate the answers of these systems against a rubric that assesses balanced perspectives, factual comprehensiveness, succinctness, and evidentiary support. Among the six systems two using RAG on curated literature outperformed existing closed models across key metrics, particularly in providing comprehensive and well-supported answers. We discuss promising aspects of LLM performances as well as critical short-comings of all the models. The set of expert-formulated questions and the rubric will be valuable for assessing expert level performance of LLM based reasoning systems.

Multimodal language models possess a remarkable ability to handle an open-vocabulary's worth of objects. Yet the best models still suffer from hallucinations when reasoning about scenes in the real world, revealing a gap between their seemingly strong performance on existing perception benchmarks that are saturating and their reasoning in the real world. To address this gap, we build a novel benchmark of in-the-wild scenes that we call Common-O. With more than 10.5k examples using exclusively new images not found in web training data to avoid contamination, Common-O goes beyond just perception, inspired by cognitive tests for humans, to probe reasoning across scenes by asking "what's in common?". We evaluate leading multimodal language models, including models specifically trained to perform chain-of-thought reasoning. We find that perceiving objects in single images is tractable for most models, yet reasoning across scenes is very challenging even for the best models, including reasoning models. Despite saturating many leaderboards focusing on perception, the best performing model only achieves 35% on Common-O -- and on Common-O Complex, consisting of more complex scenes, the best model achieves only 1%. Curiously, we find models are more prone to hallucinate when similar objects are present in the scene, suggesting models may be relying on object co-occurrence seen during training. Among the models we evaluated, we found scale can provide modest improvements while models explicitly trained with multi-image inputs show bigger improvements, suggesting scaled multi-image training may offer promise. We make our benchmark publicly available to spur research into the challenge of hallucination when reasoning across scenes.

TextGrad is a novel approach to text-based automatic differentiation that enables composite AI systems to perform optimization without explicit numerical equations. However, it currently lacks self-verification mechanisms that ensure reasoning validity in text-based decision making. This research introduces TextualVerifier, a verification framework that leverages chain-of-thought reasoning and majority voting with large language models to address this verification gap. TextualVerifier implements a four-stage workflow: chain-of-thought decomposition, variant generation, majority voting, and consensus aggregation. It integrates non-invasively with TextGrad at both the loss function and optimization result verification stages. Experimental evaluation using the Gemini 1.5 Pro model is conducted in two phases: (1) standalone evaluation on PRM800K, and (2) integrated evaluation with TextGrad on GPQA-Diamond, MMLU-ML, and MMLU-CP benchmarks. Results show statistically significant improvements (p < 0.001). In phase one, TextualVerifier improves the validity of reasoning steps by 29 percent. In phase two, integration into TextGrad loss function yields a 2.2 percentage point gain from 68.2 to 70.4 percent with a moderate overhead of 5.9 LLM calls on average. Further evaluations of TextualVerifier versioning yield 8.08, 10.71, and 3.92 percentage point improvements on GPQA, MMLU-ML, and MMLU-CP respectively. TextualVerifier thus presents the first self-verification framework for TextGrad through LLM-based techniques without requiring numerical gradients, enabling more reliable reasoning and opening new directions for verification in text-based optimization.