We introduce SciVer, the first benchmark specifically designed to evaluate the ability of foundation models to verify claims within a multimodal scientific context. SciVer consists of 3,000 expert-annotated examples over 1,113 scientific papers, covering four subsets, each representing a common reasoning type in multimodal scientific claim verification. To enable fine-grained evaluation, each example includes expert-annotated supporting evidence. We assess the performance of 21 state-of-the-art multimodal foundation models, including o4-mini, Gemini-2.5-Flash, Llama-3.2-Vision, and Qwen2.5-VL. Our experiment reveals a substantial performance gap between these models and human experts on SciVer. Through an in-depth analysis of retrieval-augmented generation (RAG), and human-conducted error evaluations, we identify critical limitations in current open-source models, offering key insights to advance models' comprehension and reasoning in multimodal scientific literature tasks.

Generating grounded and trustworthy responses remains a key challenge for large language models (LLMs). While retrieval-augmented generation (RAG) with citation-based grounding holds promise, instruction-tuned models frequently fail even in straightforward scenarios: missing explicitly stated answers, citing incorrectly, or refusing when evidence is available. In this work, we explore how reinforcement learning (RL) and internal reasoning can enhance grounding in LLMs. We use the GRPO (Group Relative Policy Optimization) method to train models using verifiable outcome-based rewards targeting answer correctness, citation sufficiency, and refusal quality, without requiring gold reasoning traces or expensive annotations. Through comprehensive experiments across ASQA, QAMPARI, ELI5, and ExpertQA we show that reasoning-augmented models significantly outperform instruction-only variants, especially in handling unanswerable queries and generating well-cited responses. A two-stage training setup, first optimizing answer and citation behavior and then refusal, further improves grounding by stabilizing the learning signal. Additionally, we revisit instruction tuning via GPT-4 distillation and find that combining it with GRPO enhances performance on long-form, generative QA tasks. Overall, our findings highlight the value of reasoning, stage-wise optimization, and outcome-driven RL for building more verifiable and reliable LLMs.

Dimensionality reduction (DR) techniques map high-dimensional data into lower-dimensional spaces. Yet, current DR techniques are not designed to explore semantic structure that is not directly available in the form of variables or class labels. We introduce a novel user-guided projection framework for image and text data that enables customizable, interpretable, data visualizations via zero-shot classification with Multimodal Large Language Models (MLLMs). We enable users to steer projections dynamically via natural-language guiding prompts, to specify high-level semantic relationships of interest to the users which are not explicitly present in the data dimensions. We evaluate our method across several datasets and show that it not only enhances cluster separation, but also transforms DR into an interactive, user-driven process. Our approach bridges the gap between fully automated DR techniques and human-centered data exploration, offering a flexible and adaptive way to tailor projections to specific analytical needs.

Text-based hyperbole and metaphor detection are of great significance for natural language processing (NLP) tasks. However, due to their semantic obscurity and expressive diversity, it is rather challenging to identify them. Existing methods mostly focus on superficial text features, ignoring the associations of hyperbole and metaphor as well as the effect of implicit emotion on perceiving these rhetorical devices. To implement these hypotheses, we propose an emotion-guided hyperbole and metaphor detection framework based on bidirectional dynamic interaction (EmoBi). Firstly, the emotion analysis module deeply mines the emotion connotations behind hyperbole and metaphor. Next, the emotion-based domain mapping module identifies the target and source domains to gain a deeper understanding of the implicit meanings of hyperbole and metaphor. Finally, the bidirectional dynamic interaction module enables the mutual promotion between hyperbole and metaphor. Meanwhile, a verification mechanism is designed to ensure detection accuracy and reliability. Experiments show that EmoBi outperforms all baseline methods on four datasets. Specifically, compared to the current SoTA, the F1 score increased by 28.1% for hyperbole detection on the TroFi dataset and 23.1% for metaphor detection on the HYPO-L dataset. These results, underpinned by in-depth analyses, underscore the effectiveness and potential of our approach for advancing hyperbole and metaphor detection.

With the popularity of large language models (LLMs), undesirable societal problems like misinformation production and academic misconduct have been more severe, making LLM-generated text detection now of unprecedented importance. Although existing methods have made remarkable progress, a new challenge posed by text from privately tuned LLMs remains underexplored. Users could easily possess private LLMs by fine-tuning an open-source one with private corpora, resulting in a significant performance drop of existing detectors in practice. To address this issue, we propose PhantomHunter, an LLM-generated text detector specialized for detecting text from unseen, privately-tuned LLMs. Its family-aware learning framework captures family-level traits shared across the base models and their derivatives, instead of memorizing individual characteristics. Experiments on data from LLaMA, Gemma, and Mistral families show its superiority over 7 baselines and 3 industrial services, with F1 scores of over 96%.

Visual Navigation is a core task in Embodied AI, enabling agents to navigate complex environments toward given objectives. Across diverse settings within Navigation tasks, many necessitate the modelling of sequential data accumulated from preceding time steps. While existing methods perform well, they typically process all historical observations simultaneously, overlooking the internal association structure within the data, which may limit the potential for further improvements in task performance. We address this by examining the unique characteristics of Navigation tasks through the lens of causality, introducing a causal framework to highlight the limitations of conventional sequential methods. Leveraging this insight, we propose Causality-Aware Navigation (CAN), which incorporates a Causal Understanding Module to enhance the agent's environmental understanding capability. Empirical evaluations show that our approach consistently outperforms baselines across various tasks and simulation environments. Extensive ablations studies attribute these gains to the Causal Understanding Module, which generalizes effectively in both Reinforcement and Supervised Learning settings without computational overhead.

Monitoring animal populations in natural environments requires systems that can interpret both visual data and human language queries. This work introduces ViLLa (Vision-Language-Logic Approach), a neuro-symbolic framework designed for interpretable animal monitoring. ViLLa integrates three core components: a visual detection module for identifying animals and their spatial locations in images, a language parser for understanding natural language queries, and a symbolic reasoning layer that applies logic-based inference to answer those queries. Given an image and a question such as "How many dogs are in the scene?" or "Where is the buffalo?", the system grounds visual detections into symbolic facts and uses predefined rules to compute accurate answers related to count, presence, and location. Unlike end-to-end black-box models, ViLLa separates perception, understanding, and reasoning, offering modularity and transparency. The system was evaluated on a range of animal imagery tasks and demonstrates the ability to bridge visual content with structured, human-interpretable queries.

Large language models excel at many tasks but still struggle with consistent, robust reasoning. We introduce Cohort-based Consistency Learning (CC-Learn), a reinforcement learning framework that improves the reliability of LLM reasoning by training on cohorts of similar questions derived from shared programmatic abstractions. To enforce cohort-level consistency, we define a composite objective combining cohort accuracy, a retrieval bonus for effective problem decomposition, and a rejection penalty for trivial or invalid lookups that reinforcement learning can directly optimize, unlike supervised fine-tuning. Optimizing this reward guides the model to adopt uniform reasoning patterns across all cohort members. Experiments on challenging reasoning benchmarks (including ARC-Challenge and StrategyQA) show that CC-Learn boosts both accuracy and reasoning stability over pretrained and SFT baselines. These results demonstrate that cohort-level RL effectively enhances reasoning consistency in LLMs.

Large Language Models (LLMs) have shown promise as decision-makers in dynamic settings, but their stateless nature necessitates creating a natural language representation of history. We present a unifying framework for systematically constructing natural language "state" representations for prompting LLM agents in repeated multi-agent games. Previous work on games with LLM agents has taken an ad hoc approach to encoding game history, which not only obscures the impact of state representation on agents' behavior, but also limits comparability between studies. Our framework addresses these gaps by characterizing methods of state representation along three axes: action informativeness (i.e., the extent to which the state representation captures actions played); reward informativeness (i.e., the extent to which the state representation describes rewards obtained); and prompting style (or natural language compression, i.e., the extent to which the full text history is summarized). We apply this framework to a dynamic selfish routing game, chosen because it admits a simple equilibrium both in theory and in human subject experiments \cite{rapoport_choice_2009}. Despite the game's relative simplicity, we find that there are key dependencies of LLM agent behavior on the natural language state representation. In particular, we observe that representations which provide agents with (1) summarized, rather than complete, natural language representations of past history; (2) information about regrets, rather than raw payoffs; and (3) limited information about others' actions lead to behavior that more closely matches game theoretic equilibrium predictions, and with more stable game play by the agents. By contrast, other representations can exhibit either large deviations from equilibrium, higher variation in dynamic game play over time, or both.

We propose a new method for estimating how much a model knows about a datapoint and use it to measure the capacity of modern language models. Prior studies of language model memorization have struggled to disentangle memorization from generalization. We formally separate memorization into two components: unintended memorization, the information a model contains about a specific dataset, and generalization, the information a model contains about the true data-generation process. When we completely eliminate generalization, we can compute the total memorization, which provides an estimate of model capacity: our measurements estimate that GPT-style models have a capacity of approximately 3.6 bits per parameter. We train language models on datasets of increasing size and observe that models memorize until their capacity fills, at which point "grokking" begins, and unintended memorization decreases as models begin to generalize. We train hundreds of transformer language models ranging from $500K$ to $1.5B$ parameters and produce a series of scaling laws relating model capacity and data size to membership inference.