We present a tendency of large language models (LLMs) to generate absurd patterns despite their clear inappropriateness in a simple task of identifying regularities in number series. Several approaches have been proposed to apply LLMs to complex real-world tasks, such as providing knowledge through retrieval-augmented generation and executing multi-step tasks using AI agent frameworks. However, these approaches rely on the logical consistency and self-coherence of LLMs, making it crucial to evaluate these aspects and consider potential countermeasures. To identify cases where LLMs fail to maintain logical consistency, we conducted an experiment in which LLMs were asked to explain the patterns in various integer sequences, ranging from arithmetic sequences to randomly generated integer series. While the models successfully identified correct patterns in arithmetic and geometric sequences, they frequently over-recognized patterns that were inconsistent with the given numbers when analyzing randomly generated series. This issue was observed even in multi-step reasoning models, including OpenAI o3, o4-mini, and Google Gemini 2.5 Flash Preview Thinking. This tendency to perceive non-existent patterns can be interpreted as the AI model equivalent of Idola Tribus and highlights potential limitations in their capability for applied tasks requiring logical reasoning, even when employing chain-of-thought reasoning mechanisms.

The relevance of this research lies in the growing demand for unmanned aerial vehicles (UAVs) capable of operating reliably in complex environments where conventional navigation becomes unreliable due to interference, poor visibility, or camouflage. Hyperspectral imaging (HSI) provides unique opportunities for UAV-based computer vision by enabling fine-grained material recognition and object differentiation, which are critical for navigation, surveillance, agriculture, and environmental monitoring. The aim of this work is to develop a deep learning architecture integrating HSI into UAV perception for navigation, object detection, and terrain classification. Objectives include: reviewing existing HSI methods, designing a hybrid 2D/3D convolutional architecture with spectral-spatial cross-attention, training, and benchmarking. The methodology is based on the modification of the Mobile 3D Vision Transformer (MDvT) by introducing the proposed SpectralCA block. This block employs bi-directional cross-attention to fuse spectral and spatial features, enhancing accuracy while reducing parameters and inference time. Experimental evaluation was conducted on the WHU-Hi-HongHu dataset, with results assessed using Overall Accuracy, Average Accuracy, and the Kappa coefficient. The findings confirm that the proposed architecture improves UAV perception efficiency, enabling real-time operation for navigation, object recognition, and environmental monitoring tasks. Keywords: SpectralCA, deep learning, computer vision, hyperspectral imaging, unmanned aerial vehicle, object detection, semi-supervised learning.

Language models often show little to no improvement (i.e., "saturation") when trained via vanilla supervised fine-tuning (SFT) on data similar to what they saw in their training set (e.g., MATH). We introduce a new fine-tuning strategy, STAT, to train such a student model by using the metacognition ability of a stronger large language model (LLM) as the teacher. The teacher uses the task dataset to create a list of skills needed for the task, and then labels each data point with its required skills (Didolkar et al., 2024). By monitoring the student's answers, the teacher creates a Missing-Skill-Profile for the student, tracking how often they failed to apply each skill in their responses. We use this idea to build a modified training set in one of two ways. In STAT-Sel, the teacher uses an existing set of training examples but adaptively reweights them according to the Missing-Skill-Profile. In STAT-Syn, the teacher synthesizes additional examples involving missing skills. Across extensive experiments on Llama and Qwen models, our methods yield improvements of up to 7.5% on MATH, whereas SFT provides only limited gains. Furthermore, STAT enhances performance on out-of-distribution benchmarks (e.g., AIME24/25, AMC23, etc.) by an average of 4.6%. Crucially, we find that STAT is complementary to RL via GRPO (Shao et al., 2024): after the model is improved using STAT to address skill gaps, GRPO continues to add further gains. We conclude that skill-targeted adaptive training should broadly improve current training pipelines. Our code is available at: https://github.com/princeton-pli/STAT.

Trust is one of the most important factors shaping whether and how people adopt and rely on artificial intelligence (AI). Yet most existing studies measure trust in terms of functionality, focusing on whether a system is reliable, accurate, or easy to use, while giving less attention to the social and emotional dimensions that are increasingly relevant for today's generative AI (GenAI) systems. These systems do not just process information; they converse, respond, and collaborate with users, blurring the line between tool and partner. In this study, we introduce and validate the Human-AI Trust Scale (HAITS), a new measure designed to capture both the rational and relational aspects of trust in GenAI. Drawing on prior trust theories, qualitative interviews, and two waves of large-scale surveys in China and the United States, we used exploratory (n = 1,546) and confirmatory (n = 1,426) factor analyses to identify four key dimensions of trust: Affective Trust, Competence Trust, Benevolence & Integrity, and Perceived Risk. We then applied latent profile analysis to classify users into six distinct trust profiles, revealing meaningful differences in how affective-competence trust and trust-distrust frameworks coexist across individuals and cultures. Our findings offer a validated, culturally sensitive tool for measuring trust in GenAI and provide new insight into how trust evolves in human-AI interaction. By integrating instrumental and relational perspectives of trust, this work lays the foundation for more nuanced research and design of trustworthy AI systems.

Fundamental questions remain about when and why adversarial examples arise in neural networks, with competing views characterising them either as artifacts of the irregularities in the decision landscape or as products of sensitivity to non-robust input features. In this paper, we instead argue that adversarial vulnerability can stem from efficient information encoding in neural networks. Specifically, we show how superposition - where networks represent more features than they have dimensions - creates arrangements of latent representations that adversaries can exploit. We demonstrate that adversarial perturbations leverage interference between superposed features, making attack patterns predictable from feature arrangements. Our framework provides a mechanistic explanation for two known phenomena: adversarial attack transferability between models with similar training regimes and class-specific vulnerability patterns. In synthetic settings with precisely controlled superposition, we establish that superposition suffices to create adversarial vulnerability. We then demonstrate that these findings persist in a ViT trained on CIFAR-10. These findings reveal adversarial vulnerability can be a byproduct of networks' representational compression, rather than flaws in the learning process or non-robust inputs.

We propose QeRL, a Quantization-enhanced Reinforcement Learning framework for large language models (LLMs). While RL is essential for LLMs' reasoning capabilities, it is resource-intensive, requiring substantial GPU memory and long rollout durations. Beyond efficiency, our findings show that quantization noise increases policy entropy, enhancing exploration, and enabling the discovery of better strategies during RL. To further optimize exploration, QeRL introduces an Adaptive Quantization Noise (AQN) mechanism, which dynamically adjusts noise during training. Experiments demonstrate that QeRL delivers over 1.5 speedup in the rollout phase. Moreover, this is the first framework to enable RL training of a 32B LLM on a single H100 80GB GPU, while delivering overall speedups for RL training. It also achieves faster reward growth and higher final accuracy than 16-bit LoRA and QLoRA, while matching the performance of full-parameter fine-tuning on mathematical benchmarks such as GSM8K (90.8%) and MA TH 500 (77.4%) in the 7B model. These results establish QeRL as an efficient and effective framework for RL training in LLMs.Figure 1: Rollout speedup and accuracy of QeRL on Qwen2.5-7B-Instruct. QeRL achieves faster RL rollout and end-to-end training speeds (batch=8), while delivering performance superior to vanilla LoRA and QLoRA, also comparable to full-parameter RL on mathematical benchmarks. The ability to perform multi-step reasoning is critical for large language models (LLMs) to handle complex tasks, from theoretical problem solving to practical decision making (Sui et al., 2025; Xu et al., 2025; Chu et al., 2025; Y ang et al., 2021). Supervised fine-tuning (SFT) is a common method to improve reasoning by training models to replicate explicit reasoning steps (Huang et al., 2024d; Min et al., 2024). In contrast, reinforcement learning (RL) uses verifiable reward signals to support adaptive learning, allowing models to explore diverse reasoning traces and identify more robust solutions (Lambert et al., 2024; DeepSeek-AI, 2025; Chen et al., 2025a). 1 AQN dynamically adjusts quantization noise with an exponential scheduler, enhancing exploration.

With the advent of DeepSeek-R1, a new wave of reinforcement learning (RL) methods has emerged that seem to unlock stronger mathematical reasoning. However, a closer look at the open-source ecosystem reveals a critical limitation: with sufficiently many draws (e.g., $\texttt{pass@1024}$), many existing base models already solve nearly all questions on widely used math benchmarks such as MATH-500 and AIME 2024. This suggests that the RL fine-tuning methods prevalent in the LLM reasoning literature largely sharpen existing solution modes rather than discovering entirely new ones. Such sharpening stands in contrast to the broader promise of RL: to foster exploration and to acquire new skills. To move beyond this plateau, we introduce MATH-Beyond (MATH-B), a benchmark deliberately constructed to defeat common open-source models of up to 8B parameters even under large sampling budgets. Improving performance on our benchmark via RL requires methods that learn to reason in ways that go beyond base model capabilities in repeated sampling. Since the problems are drawn from subsets of DAPO-Math-17K and DeepScaleR datasets, they remain topically equivalent to standard high-school math. Validating our premise, RL fine-tuned models such as Nemotron-Research-Reasoning-Qwen-1.5B and DeepScaleR-1.5B-Preview perform poorly on MATH-B at $\texttt{pass@1024}$, showing how existing approaches fall short on tackling harder instances. We hope MATH-B will catalyze exploration-driven RL approaches that elicit deeper reasoning capabilities. We release MATH-B at https://huggingface.co/datasets/brendel-group/MATH-Beyond.

In recent years, the research focus of large language models (LLMs) and agents has shifted increasingly from demonstrating novel capabilities to complex reasoning and tackling challenging tasks. However, existing evaluations focus mainly on math/code contests or general tasks, while existing multi-domain academic benchmarks lack sufficient reasoning depth, leaving the field without a rigorous benchmark for high-level reasoning. To fill this gap, we introduce the Acadreason benchmark, designed to evaluate the ability of LLMs and agents to acquire and reason over academic knowledge. It consists of 50 expert-annotated academic problems across five high-reasoning domains, including computer science, economics, law, mathematics, and philosophy. All questions are sourced from top-tier publications in recent years and undergo rigorous annotation and quality control to ensure they are both challenging and answerable. We conduct systematic evaluations of over 10 mainstream LLMs and agents. The results show that most LLMs scored below 20 points, with even the cutting-edge GPT-5 achieving only 16 points. While agents achieved higher scores, none exceeded 40 points. This demonstrates the current capability gap between LLMs and agents in super-intelligent academic research tasks and highlights the challenges of Acadreason.

Recent studies in 3D object detection for autonomous vehicles aim to enrich features through the utilization of multi-modal setups or the extraction of local patterns within LiDAR point clouds. However, multi-modal methods face significant challenges in feature alignment, and gaining features locally can be oversimplified for complex 3D object detection tasks. In this paper, we propose a novel model, NV3D, which utilizes local features acquired from voxel neighbors, as normal vectors computed per voxel basis using K-nearest neighbors (KNN) and principal component analysis (PCA). This informative feature enables NV3D to determine the relationship between the surface and pertinent target entities, including cars, pedestrians, or cyclists. During the normal vector extraction process, NV3D offers two distinct sampling strategies: normal vector density-based sampling and FOV-aware bin-based sampling, allowing elimination of up to 55% of data while maintaining performance. In addition, we applied element-wise attention fusion, which accepts voxel features as the query and value and normal vector features as the key, similar to the attention mechanism. Our method is trained on the KITTI dataset and has demonstrated superior performance in car and cyclist detection owing to their spatial shapes. In the validation set, NV3D without sampling achieves 86.60% and 80.18% mean Average Precision (mAP), greater than the baseline Voxel R-CNN by 2.61% and 4.23% mAP, respectively. With both samplings, NV3D achieves 85.54% mAP in car detection, exceeding the baseline by 1.56% mAP, despite roughly 55% of voxels being filtered out.

Current neural re-rankers often struggle with complex information needs and long, content-rich documents. The fundamental issue is not computational--it is intelligent content selection: identifying what matters in lengthy, multi-faceted texts. While humans naturally anchor their understanding around key entities and concepts, neural models process text within rigid token windows, treating all interactions as equally important and missing critical semantic signals. We introduce REGENT, a neural re-ranking model that mimics human-like understanding by using entities as a "semantic skeleton" to guide attention. REGENT integrates relevance guidance directly into the attention mechanism, combining fine-grained lexical matching with high-level semantic reasoning. This relevance-guided attention enables the model to focus on conceptually important content while maintaining sensitivity to precise term matches. REGENT achieves new state-of-the-art performance in three challenging datasets, providing up to 108% improvement over BM25 and consistently outperforming strong baselines including ColBERT and RankVicuna. To our knowledge, this is the first work to successfully integrate entity semantics directly into neural attention, establishing a new paradigm for entity-aware information retrieval.