Large vision-language models have recently demonstrated impressive performance in planning and control tasks, driving interest in their application to real-world robotics. However, deploying these models for reasoning in embodied contexts is limited by their ability to incorporate long-term experience collected across multiple days and represented by vast collections of images. Current VLMs typically struggle to process more than a few hundred images concurrently, highlighting the need for more efficient mechanisms to handle long-term memory in embodied settings. To effectively evaluate these models for long-horizon control, a benchmark must specifically target scenarios where memory is crucial for success. Existing long-video QA benchmarks overlook embodied challenges like object manipulation and navigation, which demand low-level skills and fine-grained reasoning over past interactions. Moreover, effective memory integration in embodied agents involves both recalling relevant historical information and executing actions based on that information, making it essential to study these aspects together rather than in isolation. In this work, we introduce a new benchmark for long-range embodied tasks in the Habitat simulator. This benchmark evaluates memory-based capabilities across 60 tasks requiring sustained engagement and contextual awareness in an environment. The tasks can also be procedurally extended to longer and more challenging versions, enabling scalable evaluation of memory and reasoning. We also present baselines that integrate state-of-the-art VLMs with low level navigation policies, assessing their performance on these memory-intensive tasks and highlight areas for improvement.

Prior-data fitted networks (PFNs) have recently been proposed as a promising way to train tabular foundation models. PFNs are transformers that are pre-trained on synthetic data generated from a prespecified prior distribution and that enable Bayesian inference through in-context learning. In this paper, we introduce CausalFM, a comprehensive framework for training PFN-based foundation models in various causal inference settings. First, we formalize the construction of Bayesian priors for causal inference based on structural causal models (SCMs) in a principled way and derive necessary criteria for the validity of such priors. Building on this, we propose a novel family of prior distributions using causality-inspired Bayesian neural networks that enable CausalFM to perform Bayesian causal inference in various settings, including for back-door, front-door, and instrumental variable adjustment. Finally, we instantiate CausalFM and explicitly train models to perform in-context learning in these settings. We show that CausalFM achieves competitive in-context learning performance even when compared to baselines that are specifically trained for the task at hand. In sum, our framework can be used as a general recipe to train foundation models for various causal inference settings. In contrast to the current state-of-the-art in causal inference, CausalFM offers a novel paradigm with the potential to fundamentally change how practitioners perform causal inference in medicine, economics, and other disciplines.

Audio Large Language Models (AudioLLMs) have achieved strong results in semantic tasks like speech recognition and translation, but remain limited in modeling paralinguistic cues such as emotion. Existing approaches often treat emotion understanding as a classification problem, offering little insight into the underlying rationale behind predictions. In this work, we explore emotion reasoning, a strategy that leverages the generative capabilities of AudioLLMs to enhance emotion recognition by producing semantically aligned, evidence-grounded explanations. To support this in multitask AudioLLMs, we introduce a unified framework combining reasoning-augmented data supervision, dual-encoder architecture, and task-alternating training. This approach enables AudioLLMs to effectively learn different tasks while incorporating emotional reasoning. Experiments on IEMOCAP and MELD show that our approach not only improves emotion prediction accuracy but also enhances the coherence and evidential grounding of the generated responses. Experiments on two out-of-domain datasets demonstrate the generalization capabilities of the resulting model.

As overparameterized models become increasingly prevalent, training loss alone offers limited insight into generalization performance. While smoothness has been linked to improved generalization across various settings, directly enforcing smoothness in neural networks remains challenging. To address this, we introduce Distributional Input Projection Networks (DIPNet), a novel framework that projects inputs into learnable distributions at each layer. This distributional representation induces a smoother loss landscape with respect to the input, promoting better generalization. We provide theoretical analysis showing that DIPNet reduces both local smoothness measures and the Lipschitz constant of the network, contributing to improved generalization performance. Empirically, we validate DIPNet across a wide range of architectures and tasks, including Vision Transformers (ViTs), Large Language Models (LLMs), ResNet and MLPs. Our method consistently enhances test performance under standard settings, adversarial attacks, out-of-distribution inputs, and reasoning benchmarks. We demonstrate that the proposed input projection strategy can be seamlessly integrated into existing models, providing a general and effective approach for boosting generalization performance in modern deep learning.

Large language models (LLMs) often struggle with visualization tasks like plotting diagrams, charts, where success depends on both code correctness and visual semantics. Existing instruction-tuning datasets lack execution-grounded supervision and offer limited support for iterative code correction, resulting in fragile and unreliable plot generation. We present VisCode-200K, a large-scale instruction tuning dataset for Python-based visualization and self-correction. It contains over 200K examples from two sources: (1) validated plotting code from open-source repositories, paired with natural language instructions and rendered plots; and (2) 45K multi-turn correction dialogues from Code-Feedback, enabling models to revise faulty code using runtime feedback. We fine-tune Qwen2.5-Coder-Instruct on VisCode-200K to create VisCoder, and evaluate it on PandasPlotBench. VisCoder significantly outperforms strong open-source baselines and approaches the performance of proprietary models like GPT-4o-mini. We further adopt a self-debug evaluation protocol to assess iterative repair, demonstrating the benefits of feedback-driven learning for executable, visually accurate code generation.

Large Language Model (LLM) agents are reshaping the game industry, particularly with more intelligent and human-preferable game characters. However, existing game benchmarks fall short of practical needs: they lack evaluations of diverse LLM capabilities across various game genres, studies of agentic modules crucial for complex gameplay, and fine-tuning datasets for aligning pre-trained LLMs into gaming agents. To fill these gaps, we present Orak, a foundational benchmark designed to train and evaluate LLM agents across diverse real-world video games. Unlike existing benchmarks, Orak includes 12 popular video games spanning all major genres, enabling comprehensive studies of LLM capabilities and agentic modules essential for intricate game scenarios. To support consistent evaluation of LLMs, we introduce a plug-and-play interface based on Model Context Protocol (MCP) that enables LLMs to seamlessly connect with games and manipulate agentic modules. Additionally, we propose a fine-tuning dataset, consisting of LLM gameplay trajectories across diverse game genres. Orak offers a comprehensive evaluation framework, encompassing general game score leaderboards, LLM battle arenas, and in-depth analyses of visual input state, agentic strategies, and fine-tuning effects, establishing a foundation towards building generic gaming agents. Code is available at https://github.com/krafton-ai/Orak.

Chain-of-thought (CoT) prompting enhances reasoning in large language models (LLMs) but often leads to verbose and redundant outputs, thus increasing inference cost. We hypothesize that many reasoning steps are unnecessary for producing correct answers. To investigate this, we start with a systematic study to examine what is the minimum reasoning required for a model to reach a stable decision. We find that on math reasoning tasks like math, models typically converge to their final answers after 60\% of the reasoning steps, suggesting substantial redundancy in the remaining content. Based on these insights, we propose three inference-time strategies to improve efficiency: (1) early stopping via answer consistency, (2) boosting the probability of generating end-of-reasoning signals, and (3) a supervised method that learns when to stop based on internal activations. Experiments across five benchmarks and five open-weights LLMs show that our methods significantly reduce token usage with little or no accuracy drop. In particular, on NaturalQuestions, Answer Consistency reduces tokens by over 40\% while further improving accuracy. Our work underscores the importance of cost-effective reasoning methods that operate at inference time, offering practical benefits for real-world applications.

Although Large Language Models (LLMs) have become capable reasoners, the problem of faithfulness persists: their reasoning can contain errors and omissions that are difficult to detect and that may obscure biases in model outputs. To address this issue, we introduce Semi-Structured Reasoning Models (SSRMs), which are trained to produce semi-structured representations of reasoning. SSRMs generate reasoning traces in a non-executable Pythonic syntax that names each reasoning step and marks its inputs and outputs. This structure allows SSRM traces to be automatically audited to identify reasoning flaws. We evaluate three types of audits: hand-crafted structured reasoning audits, written in a domain-specific language (DSL) implemented in Python; LLM-generated structured reasoning audits; and learned typicality audits, which apply probabilistic models over reasoning traces. We show that all of these methods can be used to effectively flag probable reasoning errors. Importantly, the auditability of SSRMs does not appear to compromise overall accuracy: in evaluation on twelve benchmarks and two model families, SSRMs demonstrate strong performance and generalizability relative to other models of comparable size.

Multimodal large language models (MLLMs) have achieved remarkable success on complex multimodal tasks. However, it remains insufficiently explored whether they exhibit modality preference, a tendency to favor one modality over another when processing multimodal contexts. Extensive experiments reveal that all 20 tested MLLMs generally demonstrate clear modality preferences, and such preferences can serve as a useful indicator of downstream task performances of MLLMs. Further analysis shows that modality preference can be controlled by instruction guidance and captured within the latent representations of MLLMs. Built on these insights, we propose a probing and steering method based on representation engineering to explicitly control modality preference without requiring additional fine-tuning. This method effectively amplifies modality preference toward a desired direction and demonstrates promising improvements across multiple downstream applications, including multimodal visual understanding and multimodal machine translation. Multimodal Large Language Models (MLLMs; Achiam et al., 2023; Team et al., 2023; Wang et al., 2024; Yin et al., 2024) have emerged as a powerful paradigm for processing and reasoning across heterogeneous data modalities (e.g., text, images, video). Recent advances demonstrate their exceptional capabilities on complex tasks with multimodal contexts, including autonomous web browsing (He et al., 2024), graphical user interface understanding (Hong et al., 2024b), and multimodal dialogue systems (Sun et al., 2022). Despite impressive performance, fundamental questions remain about their modality preference--whether MLLMs tend to rely more heavily on one modality than others, and to what extent they favor a specific modality when resolving multimodal inputs. To investigate this, one line of work (Fu et al., 2024; Amara et al., 2024) compares model performance on unimodal input, providing either only text or only image input for the same question. Another line of research analyzes the relative contributions of textual and visual context, typically by removing one modality to observe the changes of the downstream performance (Park et al., 2025) or Shapley value (Alishahi et al., 2019; Parcalabescu & Frank, 2024; 2022).

Large Language Models (LLMs) have recently been widely adopted in conversational agents. However, the increasingly long interactions between users and agents accumulate extensive dialogue records, making it difficult for LLMs with limited context windows to maintain a coherent long-term dialogue memory and deliver personalized responses. While retrieval-augmented memory systems have emerged to address this issue, existing methods often depend on single-granularity memory segmentation and retrieval. This approach falls short in capturing deep memory connections, leading to partial retrieval of useful information or substantial noise, resulting in suboptimal performance. To tackle these limits, we propose MemGAS, a framework that enhances memory consolidation by constructing multi-granularity association, adaptive selection, and retrieval. MemGAS is based on multi-granularity memory units and employs Gaussian Mixture Models to cluster and associate new memories with historical ones. An entropy-based router adaptively selects optimal granularity by evaluating query relevance distributions and balancing information completeness and noise. Retrieved memories are further refined via LLM-based filtering. Experiments on four long-term memory benchmarks demonstrate that MemGAS outperforms state-of-the-art methods on both question answer and retrieval tasks, achieving superior performance across different query types and top-K settings. \footnote{https://github.com/quqxui/MemGAS}