Accurate segmentation of floating debris on water is often compromised by surface glare and changing outdoor illumination. Polarimetric imaging offers a single-sensor route to mitigate water-surface glare that disrupts semantic segmentation of floating objects. We benchmark state-of-the-art fusion networks on PoTATO, a public dataset of polarimetric images of plastic bottles in inland waterways, and compare their performance with single-image baselines using traditional models. Our results indicate that polarimetric cues help recover low-contrast objects and suppress reflection-induced false positives, raising mean IoU and lowering contour error relative to RGB inputs. These sharper masks come at a cost: the additional channels enlarge the models increasing the computational load and introducing the risk of new false positives. By providing a reproducible, diagnostic benchmark and publicly available code, we hope to help researchers choose if polarized cameras are suitable for their applications and to accelerate related research.

Recent breakthroughs in large language models (LLMs) on reasoning tasks rely heavily on massive, high-quality datasets-typically human-annotated and thus difficult to scale. While data synthesis or distillation offers a promising alternative, existing methods struggle with inconsistent data quality and an inability to dynamically adapt to the evolving capabilities of the model, leading to suboptimal training signals. To address these limitations, we introduce Socratic-Zero, a fully autonomous framework that generates high-quality training data from minimal seed examples through the co-evolution of three agents: the Teacher, the Solver, and the Generator. The Solver continuously refines its reasoning by learning from preference feedback on both successful and failed trajectories; the Teacher adaptively crafts increasingly challenging questions based on the Solver's weaknesses; and the Generator distills the Teacher's question-design strategy to enable scalable, high-fidelity curriculum generation. This closed-loop system produces a self-improving curriculum-requiring no pre-existing tasks or labels. Remarkably, starting from only 100 seed questions, our Socratic-Solver-8B achieves an average gain of +20.2 percentage points over prior data synthesis methods across seven mathematical reasoning benchmarks (AMC23, AIME24-25, Olympiad, MATH-500, Minerva, and GSM8K), with consistent gains on both Qwen3 and GLM4 series models. Even more surprisingly, synthetic data from Socratic-Generator-32B enables student LLMs to achieve superior performance compared to other state-of-the-art (SOTA) commercial LLMs on these benchmarks, including Qwen3-235B-A22B, DeepSeek-V3.1-671B, GPT-5, Gemini-2.5-Pro, Grok-4, and Claude-4.1-Opus.

Reinforcement Learning from Human Feedback (RLHF) reward models exhibit systematic failures on longtail distributions, leading to reward hacking and misalignment. We propose a mechanistic interpretability framework that identifies specialized neural circuits responsible for rare-event processing in reward models. Drawing from recent advances showing distributed specialization for rare tokens in language models\citep{liu2025no, liu2025emergent}, we hypothesize that reward models also develop functionally distinct circuits for longtail scenarios. Our theoretical framework establishes formal connections between circuit specialization, reward generalization bounds, and longtail performance. We introduce \textbf{Circuit-Aware Reward Training (CART)}, which uses circuit analysis to guide data augmentation, regularization, and ensemble strategies. This approach provides both theoretical insights into reward model failures and practical interventions for improving longtail robustness.

The performance of large language models (LLMs) is highly sensitive to the input prompt, making prompt optimization a critical task. However, real-world application is hindered by three major challenges: (1) the black-box nature of powerful proprietary LLMs, (2) the need for high sample efficiency due to query costs, and (3) the desire for privacy-preserving collaboration among multiple users. To address these challenges simultaneously, we introduce a novel framework for sample-efficient federated prompt optimization based on multi-armed bandits (MABs). The MAB framework is uniquely suited for this problem as it is (1) inherently a black-box optimization method, (2) practically sample-efficient, and (3) enables collaborative learning with theoretically guaranteed benefit from more participating agents. We first propose the Federated Prompt Optimization via Bandits (FedPOB) algorithm, a federated variant of the Linear UCB algorithm, where agents collaborate by sharing model parameters instead of raw data. We then extend our approach to the practical setting of comparative user feedback by introducing FedPOB with Preference Feedback (FedPOB-Pref), an efficient algorithm based on federated dueling bandits. Extensive experiments demonstrate that both FedPOB and FedPOB-Pref significantly outperform existing baselines and that their performance consistently improves as more agents participate in the collaboration, validating the effectiveness of our federated approach. Large language models (LLMs) have achieved impressive performance in a variety of real-world applications (Guo et al., 2025). However, the performance of LLMs has been shown to be highly sensitive to the input prompt (Zhou et al., 2023; Lin et al., 2024b). Consequently, prompt optimization, in which we aim to find the best prompt for a task, has emerged as a critical research area. Despite its growing popularity, the widespread real-world adoption of prompt optimization is still hindered by three important challenges. The first challenge is black-box access. Some of the most powerful LLMs, such as ChatGPT and Gemini (OpenAI, 2023b; Team et al., 2023), are proprietary, black-box models that are only accessible via API queries.

Recent trends in humanoid robot control have successfully employed imitation learning to enable the learned generation of smooth, human-like trajectories from human data. While these approaches make more realistic motions possible, they are limited by the amount of available motion data, and do not incorporate prior knowledge about the physical laws governing the system and its interactions with the environment. Thus they may violate such laws, leading to divergent trajectories and sliding contacts which limit real-world stability. We address such limitations via a two-pronged learning strategy which leverages the known physics of the system and fundamental control principles. First, we encode physics priors during supervised imitation learning to promote trajectory feasibility. Second, we minimize drift at inference time by applying a proportional-integral controller directly to the generated output state. We validate our method on various locomotion behaviors for the ergoCub humanoid robot, where a physics-informed loss encourages zero contact foot velocity. Our experiments demonstrate that the proposed approach is compatible with multiple controllers on a real robot and significantly improves the accuracy and physical constraint conformity of generated trajectories.

Personalizing large language models (LLMs) to individual user preferences is a critical step beyond generating generically helpful responses. However, current personalization methods are ill-suited for new users, as they typically require either slow, resource-intensive fine-tuning or a substantial amount of pre-existing user data, creating a significant cold-start problem. To address this challenge, we introduce a new paradigm for real-time personalization by learning from online pairwise preference feedback collected during text generation. We propose T-POP (T est-Time P ersonalization with O nline P reference Feedback), a novel algorithm that synergistically combines test-time alignment with dueling bandits. Without updating the LLM parameters, T-POP steers the decoding process of a frozen LLM by learning a reward function online that captures user preferences. By leveraging dueling bandits, T-POP intelligently queries the user to efficiently balance between exploring their preferences and exploiting the learned knowledge to generate personalized text. Extensive experiments demonstrate that T-POP achieves rapid and data-efficient personalization, significantly outperforming existing baselines and showing consistent improvement with more user interactions. While large language models (LLMs) have achieved remarkable success in generating human-like text, a critical frontier remains: moving from generic, one-size-fits-all responses to deeply personalized interactions.

How reliable are single-response LLM-as-a-judge ratings without references, and can we obtain fine-grained, deterministic scores in this setting? We study the common practice of asking a judge model to assign Likert-scale scores to free-text responses and show two systematic issues: scores are unstable under sampling and poorly calibrated, leading to compression near the top of the scale and frequent ties. We then propose and evaluate Latent Judges, which derive scalar ratings from internal model signals: (i) probability-weighted scores over integer ratings, (ii) verifier-style probabilities of "yes", and (iii) linear probes trained on model activations at the rating position. Across a broad suite of pairwise and single-rating benchmarks, latent methods match or surpass standard prompting, with consistent gains on pairwise accuracy and listwise ranking relevant to Best-of-N selection. Probability-weighted scores achieve the strongest single-rating correlations, while probes recover useful signals when output logits are miscalibrated. These results indicate that latent information provides deterministic and more discriminative signals for reference-free evaluation, and can improve selection and training approaches like Best-of-$N$, multi-teacher distillation, and routing.

Long-sequence processing is a critical capability for modern large language models. However, the self-attention mechanism in the standard Transformer architecture faces severe computational and memory bottlenecks when processing long sequences. While trainable sparse attention methods offer a promising solution, existing approaches such as NSA introduce excessive extra parameters and disrupt the conventional \textit{pretrain-on-short, finetune-on-long} workflow, resulting in slow convergence and difficulty in acceleration. To overcome these limitations, we introduce dense-sparse switchable attention framework, termed as InfLLM-V2. InfLLM-V2 is a trainable sparse attention that seamlessly adapts models from short to long sequences. Specifically, InfLLM-V2 reuses dense attention parameters through parameter-free architecture modification, maintaining consistency between short and long sequence processing. Additionally, InfLLM-V2 ensures computational efficiency across all sequence lengths, by using dense attention for short inputs and smoothly transitioning to sparse attention for long sequences. To achieve practical acceleration, we further introduce an efficient implementation of InfLLM-V2 that significantly reduces the computational overhead. Our experiments on long-context understanding and chain-of-thought reasoning demonstrate that InfLLM-V2 is 4$\times$ faster than dense attention while retaining 98.1% and 99.7% of the performance, respectively. Based on the InfLLM-V2 framework, we have trained and open-sourced MiniCPM4.1 (https://huggingface.co/openbmb/MiniCPM4.1-8B), a hybrid reasoning model, providing a reproducible implementation for the research community.

ABSTRACT This paper introduces V olterra Neural Ordinary Differential Equations (VNODE), a piecewise continuous V olterra Neural Network that integrates nonlinear V olterra filtering with continuous-time neural ordinary differential equations for image classification. Drawing inspiration from the visual cortex, where discrete event processing is interleaved with continuous integration, VNODE alternates between discrete V olterra feature extraction and ODE-driven state evolution. VNODE consistently outperforms state-of-the-art models with improved computational complexity as exemplified on benchmark datasets like CIFAR-10 and Imagenet-1K. Index T erms-- Neural ODEs, V olterra Neural Networks, Image Classification, Continuous-time Models 1. INTRODUCTION Over the past decade, deep learning has transformed signal and image processing. Driven by Convolutional Neural Networks, Transformers, and their variants, it has set benchmarks in image classification, action recognition, object detection, and many other computer vision tasks [1].

In science, promotional language ('hype') is increasing and can undermine objective evaluation of evidence, impede research development, and erode trust in science. In this paper, we introduce the task of automatic detection of hype, which we define as hyperbolic or subjective language that authors use to glamorize, promote, embellish, or exaggerate aspects of their research. We propose formalized guidelines for identifying hype language and apply them to annotate a portion of the National Institutes of Health (NIH) grant application corpus. We then evaluate traditional text classifiers and language models on this task, comparing their performance with a human baseline. Our experiments show that formalizing annotation guidelines can help humans reliably annotate candidate hype adjectives and that using our annotated dataset to train machine learning models yields promising results. Our findings highlight the linguistic complexity of the task, and the potential need for domain knowledge and temporal awareness of the facts. While some linguistic works address hype detection, to the best of our knowledge, we are the first to approach it as a natural language processing task.