Accurate prediction of surgical duration is pivotal for hospital resource management. Although recent supervised learning approaches, from machine learning (ML) to fine-tuned large language models (LLMs), have shown strong performance, they remain constrained by the need for high-quality labeled data and computationally intensive training. In contrast, zero-shot LLM inference offers a promising training-free alternative, but it lacks grounding in institution-specific clinical evidence, making its predictions clinically misaligned and prone to instability. To address these limitations, we present PREBA, a retrieval-augmented framework that integrates PCA-weighted retrieval and Bayesian averaging to ground LLM reasoning in institution-specific evidence without requiring task-specific fine-tuning. Specifically, PREBA first encodes heterogeneous clinical features into a unified representation. It then performs PCA-weighted retrieval, weighting each dimension by its contribution to the variation across institutional cases, instead of treating them as equally informative, and conditions the LLM to generate multiple candidate predictions. Finally, Bayesian averaging consolidates these predictions with population-level statistical priors, yielding a calibrated estimate that remains stable and clinically plausible. We evaluate PREBA on four real-world datasets spanning different institutions and languages, using three state-of-the-art LLMs, including Qwen3, DeepSeek-R1, and HuatuoGPT-o1. PREBA consistently outperforms zero-shot inference, reducing MAE by up to 64.9% and raising R2 to as high as 0.63. It further attains accuracy competitive with, and on several datasets superior to, fully supervised ML methods, while offering greater interpretability through retrieved exemplars and explicit priors. The code is available at https://github.com/kanxueli/PREBA.
Mordell's problem asks for the largest product of squared pairwise distances among $n$ points in the complex plane with a prescribed sum of squared distances from the origin. When this sum equals $n$, the maximum for $2\le n\le5$ is known to be $n^n$, attained precisely by the $n$th roots of unity up to rotation and relabeling. For $n\ge6$, Chalk showed that a regular polygon together with its center gives a larger product than the regular $n$-gon. We prove that for $6\le n\le9$ the maximum is $(n-1)^{n-1}r_n^{n(n-1)}$, where $r_n=\sqrt{n/(n-1)}$. Up to the same symmetries, the unique maximizer consists of the origin and the vertices of a regular $(n-1)$-gon of radius $r_n$. For six points, algebraic elimination reduces the proof to an exact comparison of stationary values. For seven points, we compare all twelve classes in a published classification of equilibria. For eight and nine points, rigorous interval arithmetic establishes an inequality for every admissible pair. Summing over pairs gives the sharp global bound, and the equality conditions determine the configuration. An explicit ten-point comparison and a known instability result together exclude Chalk's configuration for every $n\ge10$. For each $n$ from two through eight, Mordell's problem and the minimum-diameter problem studied by Bateman and Erdős have a common optimal configuration.
Zhenxi Liu, Guanze Chen, Dehui Yu, Xingyu Wang, Bingxu Hou, Chengwei Li, Tao Ma, Ke Wang 等 9 位作者
Image-text corpora contain copies of one image, captions that describe one image in different ways, captions reused for different images, and captions that do not match their images. Each case calls for a different curation action. Deduplication and multimodal entity resolution (ER) only decide whether records are the same, and they are evaluated on labeled pairs. Yet pair scores do not tell which records end up in the same cluster or which useful captions get deleted. We therefore formulate image-text deduplication as graded multimodal ER, which decides visual identity, caption handling, and mismatches for each pair, and we evaluate it at three levels: pairs, clusters, and the retained set. Our method, VL-Dedup, groups records of the same visual entity into clusters and caps the size of each cluster. Within each cluster, a vision-language model (VLM) and CLIPScore decide which captions to keep and which to flag. We also build VL-DedupBench, with 50,000 COCO and SOP records and 66,788 labeled pairs. On it, a classifier that selects no false edge among the labeled pairs still joins millions of record pairs from different entities. Spending VLM calls on the pairs inside clusters, rather than on pairs with uncertain scores, gives better caption decisions from the same model. Even when almost all clusters are correct, keeping one record per cluster loses facts in many deletions, while the deletions of VL-Dedup lose much less information.
Hao Jiang, Peiru Du, Pengfei Yao, Mengting Li, Siyuan Lou, Kuo Cai, Sheng Yu, Qiang Luo 等 13 位作者
Large language models (LLMs) have demonstrated strong reasoning capabilities, motivating their use as the backbone of foundation recommendation models (FRMs). Existing methods enhance recommendations through explicit Chain-of-Thought (CoT) reasoning under a Think-then-Answer paradigm. However, explicit CoT incurs substantial inference overhead by generating lengthy reasoning traces and relies on manually designed templates that struggle to capture diverse, dynamic user interests. We propose OneLatent, an efficient latent reasoning framework that compresses explicit reasoning traces into several learnable latent tokens, enabling Latent-Reason-then-Answer inference without generating verbose traces. OneLatent first introduces Multi-View Adaptive CoT (MV-ACoT), which creates diverse, high-quality teacher-generated supervision by exploring user interests from multiple perspectives and automatically adapting reasoning complexity to each instance. Building on pretrained FRMs, it then uses a three-stage latent-token alignment paradigm to progressively internalize CoT traces into learnable latent tokens. Finally, a multistage curriculum-based post-training strategy activates latent-token reasoning for downstream recommendation tasks. Experiments on an industrial-scale Kuaishou dataset and the public Kuaishou LLM-Rec benchmark show that OneLatent consistently outperforms explicit CoT-based methods and traditional baselines. Compared with the Think and No-Think variants of FRMs, OneLatent improves SID@64 by 17.44% and 9.33%, respectively, while achieving over 17x higher online inference throughput. We further develop a production serving system for scalable, real-time FRM inference. An online A/B test in Kuaishou's local-services advertising scenario shows that deploying OneLatent with this system yields an estimated 9.6% revenue lift over strong online baselines, including OneRec and OneReason.
Mathematical formalization plays an important role in establishing rigorous and machine-verifiable foundations for mathematical reasoning. With the development of large language models (LLMs) and interactive theorem provers, intelligent agents offer a promising approach to automating the formalization of mathematical theories. This paper presents the Math Foundation Formalization Agent (MFF-Agent), a general-purpose intelligent agent designed to support formalization tasks across multiple mathematical domains. The agent integrates the reasoning capabilities of large language models with the formal verification capabilities of the Rocq proof assistant. It incorporates task classification, tool-assisted execution, verification feedback, and state-based task management to support the generation, refinement, and verification of formal proofs. Unlike an approach focused on a single mathematical theory, the agent is examined through case studies involving Morse--Kelley axiomatic set theory (MK), real numbers, propositional logic, and equivalences of the axiom of choice. These cases investigate its applicability across mathematical foundations and distinguish source-level completion, interactive verification, and reproducible compilation. The results demonstrate the feasibility of a unified operational framework for heterogeneous foundation-level tasks and clarify the gap between such a framework and a fully integrated mathematical foundation.
Conventional recommender systems rely on fixed cascaded pipelines, limiting their ability to adapt to evolving user needs and feedback. Existing agentic recommenders introduce reasoning and tool use but often lack end-to-end decision control and systematic failure reflection. We propose EvoRec, a unified agentic recommendation framework that combines autonomous decision-making with reflection-driven self-evolution. A Need Agent infers user needs from long-term, short-term, and working memory. A Planning Agent adaptively allocates retrieval budgets between Behavior Continuation and Need Satisfaction, selects retrieval tools, and sets ranking-fusion weights. An Acting Executor executes the plan and records decision traces. Following user feedback, a Dynamic Reflection Tree identifies failures in need inference, planning, or previous repairs, proposes personalized corrections, and validates them through replay. Verified repairs update working memory and are distilled into reusable cross-user strategies, enabling continual improvement without parameter updates. Experiments on a Kuaishou local-services advertising dataset show a 15.7% Hit@50 improvement over the strongest baseline. Hierarchical reflection further improves Hit@50 by 13.1% among users with prior repairs. In a seven-day online A/B test, nearline EvoRec increases advertising revenue by 4.6% and CTCVR by 3.5%, with a 14.5% revenue gain for high-intent users. These results demonstrate the effectiveness of reflection-driven self-evolution in industrial-scale recommendation.
Methods for improving knowledge use in large language models typically fall into two regimes. Non-parametric retrieval offers flexible access to external knowledge, but adds retrieval latency, context overhead, and only shallow integration with the backbone. Parametric adaptation is efficient at inference time, but entangles knowledge with model weights and can be hard to update, audit, or transfer. Engram-style hashed memory occupies a middle regime: it stores learned information in an external, addressable table, yet consumes that table through a small learned reader. This raises a basic question: when such a memory is moved across backbones, what matters more, the frozen memory itself or the target-side reader? We study this question through cross-model frozen-memory extraction, in which a memory trained on a source model is frozen and attached to a different target model, with only a lightweight reader trained. Ablations show that learned memory content and correct addressing both matter, while the amount of usable signal depends on how well the reader is aligned to the target model. In downstream question answering tasks, a dual-layer, four-branch reader nearly closes the gap between same-model and cross-model reuse, achieving an average score of 38.8 under our controlled evaluation protocol. Moreover, when the provider reader is directly compatible with the target interface, the frozen artifact can provide substantial utility without target-side training, while optional reader adaptation yields further improvement. These results suggest that Engram can serve as a reusable external knowledge artifact, provided that the target has access to a compatible reader interface; target-side adaptation can further improve alignment when direct reader reuse is insufficient.
Yuquan Bi, Baosheng Yu, Yingke Lei, Jianwei Yang, Hongsong Wang, Jie Gui, Yuan Yan Tang, James Tin-Yau Kwok
High-capacity watermarking is necessary for 3D Gaussian Splatting (3DGS) assets to embed rich information (\eg, ownership, provenance, and authentication codes), enabling reliable identification and provenance tracking in large-scale 3D asset pipelines. Existing bit-to-token watermarking methods based on a pre-trained text encoder are limited to 77-bit messages due to CLIP's fixed 77-token context length, as tokens beyond this limit are unsupported by learned positional embeddings. To address this limitation, we introduce BitC-3DGS, a bit-compression framework that encodes multiple message bits per token. It employs a bit-compressed tokenization scheme that encodes multiple bits within the same chunk into a single semantic token. To enable recovery of the compressed information, it further introduces a dual-branch architecture for joint chunk decompression and bit decoding, along with a hard-message sampling strategy to improve combinatorial coverage during decoder training. Extensive experiments on the Blender and LLFF datasets demonstrate the effectiveness of BitC-3DGS for high-capacity watermarking, achieving high message recovery accuracy and rendering fidelity. For example, it supports 128-bit message capacity with recovery accuracy comparable to that of 64-bit messages in recent state-of-the-art methods.
Multi-path reasoning methods such as self-consistency (SC) sample $K$ reasoning paths and choose the most frequent answer. However, their gains quickly plateau as $K$ increases, and locating this saturation point currently requires fitting a scaling model to accuracies measured at several values of $K$ or stopping online per query. We formalize multi-path LLM reasoning as a diversity combining problem from wireless communications: each path is a noisy channel observation, and the pairwise correlation of path correctness caps the design-effect effective sample size of the vote at a finite ceiling. Generalized least squares (GLS) analysis shows that, for paths with equal gains, the optimal linear combiner of latent embeddings is uniform exactly when every row of the path covariance has the same sum, as under exchangeability, supporting majority vote as the natural default in standard SC while leaving room for weighting or pruning under heterogeneous prompt-template branches. Across 5 models and 12 benchmarks, prompt-template diversity reduces path correlation in $55$ of $60$ model-benchmark cells, with a mean relative reduction of $16\%$ and reductions in every task domain. We derive an Adaptive-K rule that uses a four-path pilot to select $K^*$, retaining $96$--$103\%$ of MV@$K{=}32$ accuracy across Math, QA, and NLU.
Zheng Zhang, Yifan Wu, Yuxiang Zhou, Qi Liu, Qingyang Mao, Ning Li, Weibo Gao, Qiyuan Shi 等 10 位作者
Large language models (LLMs) are increasingly evaluated for fairness, yet most existing benchmarks audit one sensitive attribute at a time. This single-attribute paradigm can miss intersectional harms, where multiple attributes interact non-additively and produce failures that are invisible under marginal evaluation. We introduce ComFair, a configurable benchmark that converts existing fairness data into 479,066 controlled Dual- and Multi-attribute instances and pairs every augmented instance with its task-equivalent source. Across Recognition, Selection, Continuation, and Conversation, this pairing supports a unified audit of performance or bias-score shifts, how often added attributes introduce new errors or mask existing ones, and model-ranking stability. Experiments on ten LLMs show task- and subgroup-dependent sensitivity: discriminative tasks expose both transition directions even when aggregate changes are modest, generative tasks exhibit larger and directionally asymmetric shifts, and rankings vary with attribute composition. ComFair therefore provides a scalable controlled sensitivity audit---rather than a model of the natural joint distribution of identities---for testing the robustness of fairness-benchmark conclusions beyond single-attribute settings.
Zheng Zhang, Qi Liu University of Science and Technology of China, Siyuan Liang Nanyang Technological University, Ning Li, Zirui Hu Nanyang Technological University, Weibo Gao University of Science and Technology of China, Rui Li, Zhenya Huang University of Science and Technology of China 等 11 位作者
Large language models (LLMs) have achieved remarkable performance across diverse tasks, largely driven by large-scale pretraining. However, this data abundance introduces test data contamination, where benchmark datasets overlap with pretraining corpora, undermining the reliability of model evaluation by confounding memorization with genuine generalization. To mitigate this issue, existing training data detectors attempt to identify clean (unseen) samples from contaminated test sets, but often suffer from residual contamination due to the black-box nature of LLMs. As a result, contaminated data may be mistakenly retained, leading to unreliable evaluation.To address this challenge, we propose FTD (FDR-controlled Training Data detection), a principled framework that detects and filters contaminated evaluation data while providing a statistical guarantee: the proportion of contaminated samples mistakenly retained as clean, the false discovery rate (FDR), is provably controlled below a user-specified threshold. FTD combines multiple complementary detectors via an adaptive weighting strategy, and we theoretically show it achieves high statistical power under valid FDR control. Extensive experiments on real-world benchmarks demonstrate that FTD significantly reduces residual contamination compared to existing methods while preserving evaluation consistency.
尚无 AI 解读，展示原文摘要
cs.CL
Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers) · 2026 · 正式发表