Although large language models (LLMs) are becoming increasingly capable of solving challenging real-world tasks, accurately quantifying their uncertainty remains a critical open problem, which limits their applicability in high-stakes domains. This challenge is further compounded by the closed-source, black-box nature of many state-of-the-art LLMs. Moreover, LLM-based systems can be highly sensitive to the prompts that bind them together, which often require significant manual tuning (i.e., prompt engineering). In this work, we address these challenges by viewing LLM-based systems through a Bayesian lens. We interpret prompts as textual parameters in a statistical model, allowing us to use a small training dataset to perform Bayesian inference over these prompts. This novel perspective enables principled uncertainty quantification over both the model's textual parameters and its downstream predictions, while also incorporating prior beliefs about these parameters expressed in free-form text. To perform Bayesian inference, a difficult problem even for well-studied data modalities, we introduce Metropolis-Hastings through LLM Proposals (MHLP), a novel Markov chain Monte Carlo (MCMC) algorithm that combines prompt optimization techniques with standard MCMC methods. MHLP is a turnkey modification to existing LLM pipelines, including those that rely exclusively on closed-source models. Empirically, we demonstrate that our method yields improvements in both predictive accuracy and uncertainty quantification (UQ) on a range of LLM benchmarks and UQ tasks. More broadly, our work demonstrates a viable path for incorporating methods from the rich Bayesian literature into the era of LLMs, paving the way for more reliable and calibrated LLM-based systems.

Test-time scaling paradigms have significantly advanced the capabilities of large language models (LLMs) on complex tasks. Despite their empirical success, theoretical understanding of the sample efficiency of various test-time strategies -- such as self-consistency, best-of-$n$, and self-correction -- remains limited. In this work, we first establish a separation result between two repeated sampling strategies: self-consistency requires $Θ(1/Δ^2)$ samples to produce the correct answer, while best-of-$n$ only needs $Θ(1/Δ)$, where $Δ< 1$ denotes the probability gap between the correct and second most likely answers. Next, we present an expressiveness result for the self-correction approach with verifier feedback: it enables Transformers to simulate online learning over a pool of experts at test time. Therefore, a single Transformer architecture can provably solve multiple tasks without prior knowledge of the specific task associated with a user query, extending the representation theory of Transformers from single-task to multi-task settings. Finally, we empirically validate our theoretical results, demonstrating the practical effectiveness of self-correction methods.

Improving cultural competence of language technologies is important. However most recent works rarely engage with the communities they study, and instead rely on synthetic setups and imperfect proxies of culture. In this work, we take a human-centered approach to discover and measure language-based cultural norms, and cultural competence of LLMs. We focus on a single kind of culture, research cultures, and a single task, adapting writing across research cultures. Through a set of interviews with interdisciplinary researchers, who are experts at moving between cultures, we create a framework of structural, stylistic, rhetorical, and citational norms that vary across research cultures. We operationalise these features with a suite of computational metrics and use them for (a) surfacing latent cultural norms in human-written research papers at scale; and (b) highlighting the lack of cultural competence of LLMs, and their tendency to homogenise writing. Overall, our work illustrates the efficacy of a human-centered approach to measuring cultural norms in human-written and LLM-generated texts.

Reinforcement Learning Finetuning (RFT) has significantly advanced the reasoning capabilities of large language models (LLMs) by enabling long chains of thought, self-correction, and effective tool use. While recent works attempt to extend RFT to vision-language models (VLMs), these efforts largely produce text-only reasoning conditioned on static image inputs, falling short of true multimodal reasoning in the response. In contrast, test-time methods like Visual Sketchpad incorporate visual steps but lack training mechanisms. We introduce VTool-R1, the first framework that trains VLMs to generate multimodal chains of thought by interleaving text and intermediate visual reasoning steps. VTool-R1 integrates Python-based visual editing tools into the RFT process, enabling VLMs to learn when and how to generate visual reasoning steps that benefit final reasoning. Trained with outcome-based rewards tied to task accuracy, our approach elicits strategic visual tool use for reasoning without relying on process-based supervision. Experiments on structured visual question answering over charts and tables show that VTool-R1 enhances reasoning performance by teaching VLMs to "think with images" and generate multimodal chain of thoughts with tools.

Large reasoning models (LRMs) have demonstrated impressive reasoning capabilities across a broad range of tasks including Olympiad-level mathematical problems, indicating evidence of their complex reasoning abilities. While many reasoning benchmarks focus on the STEM domain, the ability of LRMs to reason correctly in broader task domains remains underexplored. In this work, we introduce \textbf{TTT-Bench}, a new benchmark that is designed to evaluate basic strategic, spatial, and logical reasoning abilities in LRMs through a suite of four two-player Tic-Tac-Toe-style games that humans can effortlessly solve from a young age. We propose a simple yet scalable programmatic approach for generating verifiable two-player game problems for TTT-Bench. Although these games are trivial for humans, they require reasoning about the intentions of the opponent, as well as the game board's spatial configurations, to ensure a win. We evaluate a diverse set of state-of-the-art LRMs, and \textbf{discover that the models that excel at hard math problems frequently fail at these simple reasoning games}. Further testing reveals that our evaluated reasoning models score on average $\downarrow$ 41\% \& $\downarrow$ 5\% lower on TTT-Bench compared to MATH 500 \& AIME 2024 respectively, with larger models achieving higher performance using shorter reasoning traces, where most of the models struggle on long-term strategic reasoning situations on simple and new TTT-Bench tasks.

Code generation is one of the most active areas of application of Large Language Models (LLMs). While LLMs lower barriers to writing code and accelerate development process, the overall quality of generated programs depends on the quality of given prompts. Specifically, functionality and quality of generated code can be sensitive to user's background and familiarity with software development. It is therefore important to quantify LLM's sensitivity to variations in the input. To this end we propose a synthetic evaluation pipeline for code generation with LLMs, as well as a systematic persona-based evaluation approach to expose qualitative differences of LLM responses dependent on prospective user background. Both proposed methods are completely independent from specific programming tasks and LLMs, and thus are widely applicable. We provide experimental evidence illustrating utility of our methods and share our code for the benefit of the community.

Vision-and-Language Navigation (VLN) presents a complex challenge in embodied AI, requiring agents to interpret natural language instructions and navigate through visually rich, unfamiliar environments. Recent advances in large vision-language models (LVLMs), such as CLIP and Flamingo, have significantly improved multimodal understanding but introduced new challenges related to computational cost and real-time deployment. In this project, we propose a modular, plug-and-play navigation framework that decouples vision-language understanding from action planning. By integrating a frozen vision-language model, Qwen2.5-VL-7B-Instruct, with lightweight planning logic, we aim to achieve flexible, fast, and adaptable navigation without extensive model fine-tuning. Our framework leverages prompt engineering, structured history management, and a two-frame visual input strategy to enhance decision-making continuity across navigation steps. W e evaluate our system on the Room-to-Room benchmark within the VLN-CE setting using the Matterport3D dataset and Habitat-Lab simulation environment. Although our initial results reveal challenges in generalizing to unseen environments under strict evaluation settings, our modular approach lays a foundation for scalable and efficient navigation systems, highlighting promising directions for future improvement through enhanced environmental priors and expanded mul-timodal input integration.

Story generation has been a prominent application of Large Language Models (LLMs). However, understanding LLMs' ability to produce high-quality stories remains limited due to challenges in automatic evaluation methods and the high cost and subjectivity of manual evaluation. Computational narratology offers valuable insights into what constitutes a good story, which has been applied in the symbolic narrative planning approach to story generation. This work aims to deepen the understanding of LLMs' story generation capabilities by using them to solve narrative planning problems. We present a benchmark for evaluating LLMs on narrative planning based on literature examples, focusing on causal soundness, character intentionality, and dramatic conflict. Our experiments show that GPT-4 tier LLMs can generate causally sound stories at small scales, but planning with character intentionality and dramatic conflict remains challenging, requiring LLMs trained with reinforcement learning for complex reasoning. The results offer insights on the scale of stories that LLMs can generate while maintaining quality from different aspects. Our findings also highlight interesting problem solving behaviors and shed lights on challenges and considerations for applying LLM narrative planning in game environments.

Recently, large language models have shown remarkable reasoning capabilities through long-chain reasoning before responding. However, how to extend this capability to visual reasoning tasks remains an open challenge. Existing multimodal reasoning approaches transfer such visual reasoning task into textual reasoning task via several image-to-text conversions, which often lose critical structural and semantic information embedded in visualizations, especially for tasks like chart question answering that require a large amount of visual details. To bridge this gap, we propose ChartReasoner, a code-driven novel two-stage framework designed to enable precise, interpretable reasoning over charts. We first train a high-fidelity model to convert diverse chart images into structured ECharts codes, preserving both layout and data semantics as lossless as possible. Then, we design a general chart reasoning data synthesis pipeline, which leverages this pretrained transport model to automatically and scalably generate chart reasoning trajectories and utilizes a code validator to filter out low-quality samples. Finally, we train the final multimodal model using a combination of supervised fine-tuning and reinforcement learning on our synthesized chart reasoning dataset and experimental results on four public benchmarks clearly demonstrate the effectiveness of our proposed ChartReasoner. It can preserve the original details of the charts as much as possible and perform comparably with state-of-the-art open-source models while using fewer parameters, approaching the performance of proprietary systems like GPT-4o in out-of-domain settings.

The advent of Open Radio Access Networks (O-RAN) has transformed the telecommunications industry by promoting interoperability, vendor diversity, and rapid innovation. However, its disaggregated architecture introduces complex testing challenges, particularly in validating multi-vendor components against O-RAN ALLIANCE and 3GPP specifications. Existing frameworks, such as those provided by Open Testing and Integration Centres (OTICs), rely heavily on manual processes, are fragmented and prone to human error, leading to inconsistency and scalability issues. To address these limitations, we present AI5GTest -- an AI-powered, specification-aware testing framework designed to automate the validation of O-RAN components. AI5GTest leverages a cooperative Large Language Models (LLM) framework consisting of Gen-LLM, Val-LLM, and Debug-LLM. Gen-LLM automatically generates expected procedural flows for test cases based on 3GPP and O-RAN specifications, while Val-LLM cross-references signaling messages against these flows to validate compliance and detect deviations. If anomalies arise, Debug-LLM performs root cause analysis, providing insight to the failure cause. To enhance transparency and trustworthiness, AI5GTest incorporates a human-in-the-loop mechanism, where the Gen-LLM presents top-k relevant official specifications to the tester for approval before proceeding with validation. Evaluated using a range of test cases obtained from O-RAN TIFG and WG5-IOT test specifications, AI5GTest demonstrates a significant reduction in overall test execution time compared to traditional manual methods, while maintaining high validation accuracy.