Searcharxiv⌕ Search

arXiv subjects

Yang Zhou

Publications and source records attributed to Yang Zhou.

At least 163 records · Page 9Linked to original sources
arXiv2026

A boundedness theorem for principal bundles on curves

Let $G$ be a reductive group acting on an affine scheme $V$. We study the set of principal $G$-bundles on a smooth projective curve $\mathcal C$ such that the associated $V$-bundle admits a section sending the generic point of $\mathcal C$ into the GIT stable locus $V^{\mathrm{s}}(θ)$. We show that after fixing the degree of the line bundle induced by the character $θ$, the set of such principal $G$-bundles is bounded. The statement of our theorem is made slightly more general so that we deduce from it the boundedness for $ε$-stable quasimaps and $Ω$-stable LG-quasimap.

math.AG↗
arXiv2026

Jackpot: Optimal Budgeted Rejection Sampling for Extreme Actor-Policy Mismatch Reinforcement Learning

Reinforcement learning (RL) for large language models (LLMs) remains expensive, particularly because the rollout is expensive. Decoupling rollout generation from policy optimization (e.g., leveraging a more efficient model to rollout) could enable substantial efficiency gains, yet doing so introduces a severe distribution mismatch that destabilizes learning. We propose Jackpot, a framework that leverages Optimal Budget Rejection Sampling (OBRS) to directly reduce the discrepancy between the rollout model and the evolving policy. Jackpot integrates a principled OBRS procedure, a unified training objective that jointly updates the policy and rollout models, and an efficient system implementation enabled by top-$k$ probability estimation and batch-level bias correction. Our theoretical analysis shows that OBRS consistently moves the rollout distribution closer to the target distribution under a controllable acceptance budget. Empirically, \sys substantially improves training stability compared to importance-sampling baselines, achieving performance comparable to on-policy RL when training Qwen3-8B-Base for up to 300 update steps of batchsize 64. Taken together, our results show that OBRS-based alignment brings us a step closer to practical and effective decoupling of rollout generation from policy optimization for RL for LLMs.

cs.AI↗
arXiv2026

M^3-Bench: Multi-Modal, Multi-Hop, Multi-Threaded Tool-Using MLLM Agent Benchmark

We present M^3-Bench, the first benchmark for evaluating multimodal tool use under the Model Context Protocol. The benchmark targets realistic, multi-hop and multi-threaded workflows that require visual grounding and textual reasoning, cross-tool dependencies, and persistence of intermediate resources across steps. We introduce a similarity-driven alignment that serializes each tool call, embeds signatures with a sentence encoder, and performs similarity-bucketed Hungarian matching to obtain auditable one-to-one correspondences. On top of this alignment, we report interpretable metrics that decouple semantic fidelity from workflow consistency. The benchmark spans 28 servers with 231 tools, and provides standardized trajectories curated through an Executor & Judge pipeline with human verification; an auxiliary four large language models (LLMs) judge ensemble reports end-task Task Completion and information grounding. Evaluations of representative state-of-the-art Multimodal LLMs (MLLMs) reveal persistent gaps in multimodal MCP tool use, particularly in argument fidelity and structure consistency, underscoring the need for methods that jointly reason over images, text, and tool graphs. Our Benchmark's anonymous repository is at https://github.com/EtaYang10th/Open-M3-Bench

cs.AI↗
arXiv2026

OpenOneRec Technical Report

While the OneRec series has successfully unified the fragmented recommendation pipeline into an end-to-end generative framework, a significant gap remains between recommendation systems and general intelligence. Constrained by isolated data, they operate as domain specialists-proficient in pattern matching but lacking world knowledge, reasoning capabilities, and instruction following. This limitation is further compounded by the lack of a holistic benchmark to evaluate such integrated capabilities. To address this, our contributions are: 1) RecIF Bench & Open Data: We propose RecIF-Bench, a holistic benchmark covering 8 diverse tasks that thoroughly evaluate capabilities from fundamental prediction to complex reasoning. Concurrently, we release a massive training dataset comprising 96 million interactions from 160,000 users to facilitate reproducible research. 2) Framework & Scaling: To ensure full reproducibility, we open-source our comprehensive training pipeline, encompassing data processing, co-pretraining, and post-training. Leveraging this framework, we demonstrate that recommendation capabilities can scale predictably while mitigating catastrophic forgetting of general knowledge. 3) OneRec-Foundation: We release OneRec Foundation (1.7B and 8B), a family of models establishing new state-of-the-art (SOTA) results across all tasks in RecIF-Bench. Furthermore, when transferred to the Amazon benchmark, our models surpass the strongest baselines with an average 26.8% improvement in Recall@10 across 10 diverse datasets (Figure 1). This work marks a step towards building truly intelligent recommender systems. Nonetheless, realizing this vision presents significant technical and theoretical challenges, highlighting the need for broader research engagement in this promising direction.

cs.IR↗
arXiv2026

Unveiling Traffic Wave of Linear Adaptive Cruise Control: A Second-order Macroscopic Traffic Flow Model

Traffic waves, the spatiotemporal propagation of congestion, are a key feature of traffic flow. As Adaptive Cruise Control (ACC) systems gain widespread adoption and show promise for improving both efficiency and safety, understanding how these waves evolve under ACC becomes increasingly important. Yet most existing analyses rely on steady-state metrics (e.g., equilibrium spacing) and neglect the ACC control-law parameters, such as feedback gains, that fundamentally shape higher-order traffic dynamics. To overcome this limitation, we embed the ACC control law directly into the momentum equation while retaining mass conservation law. The result is a higher-order macroscopic model whose dynamics are governed by a second-order partial differential equation equivalent to the linear ACC feedback law. Analyzing the flux Jacobian confirms that the system is strictly hyperbolic, thereby preserving anisotropy and ensuring physical consistency. The derivation also shows that traffic wave evolution depends on both the initial state and the ACC control parameters. We analyze wave-propagation characteristics, linear degeneracy, admissible discontinuities, and their connection to ACC string stability, with the corresponding derivations. Numerical experiments confirm that the second-order model yields markedly lower vehicle-pair speed deviations along wave paths than a first-order model subject to the same non-steady disturbances, underscoring both the necessity of a second-order treatment and the soundness of the proposed framework.

math.AP↗
arXiv2026

Unrewarded Exploration in Large Language Models Reveals Latent Learning from Psychology

Latent learning, classically theorized by Tolman, shows that biological agents (e.g., rats) can acquire internal representations of their environment without rewards, enabling rapid adaptation once rewards are introduced. In contrast, from a cognitive science perspective, reward learning remains overly dependent on external feedback, limiting flexibility and generalization. Although recent advances in the reasoning capabilities of large language models (LLMs), such as OpenAI-o1 and DeepSeek-R1, mark a significant breakthrough, these models still rely primarily on reward-centric reinforcement learning paradigms. Whether and how the well-established phenomenon of latent learning in psychology can inform or emerge within LLMs' training remains largely unexplored. In this work, we present novel findings from our experiments that LLMs also exhibit the latent learning dynamics. During an initial phase of unrewarded exploration, LLMs display modest performance improvements, as this phase allows LLMs to organize task-relevant knowledge without being constrained by reward-driven biases, and performance is further enhanced once rewards are introduced. LLMs post-trained under this two-stage exploration regime ultimately achieve higher competence than those post-trained with reward-based reinforcement learning throughout. Beyond these empirical observations, we also provide theoretical analyses for our experiments explaining why unrewarded exploration yields performance gains, offering a mechanistic account of these dynamics. Specifically, we conducted extensive experiments across multiple model families and diverse task domains to establish the existence of the latent learning dynamics in LLMs.

cs.LG↗
arXiv2026

Note2Chat: Improving LLMs for Multi-Turn Clinical History Taking Using Medical Notes

Effective clinical history taking is a foundational yet underexplored component of clinical reasoning. While large language models (LLMs) have shown promise on static benchmarks, they often fall short in dynamic, multi-turn diagnostic settings that require iterative questioning and hypothesis refinement. To address this gap, we propose \method{}, a note-driven framework that trains LLMs to conduct structured history taking and diagnosis by learning from widely available medical notes. Instead of relying on scarce and sensitive dialogue data, we convert real-world medical notes into high-quality doctor-patient dialogues using a decision tree-guided generation and refinement pipeline. We then propose a three-stage fine-tuning strategy combining supervised learning, simulated data augmentation, and preference learning. Furthermore, we propose a novel single-turn reasoning paradigm that reframes history taking as a sequence of single-turn reasoning problems. This design enhances interpretability and enables local supervision, dynamic adaptation, and greater sample efficiency. Experimental results show that our method substantially improves clinical reasoning, achieving gains of +16.9 F1 and +21.0 Top-1 diagnostic accuracy over GPT-4o. Our code and dataset can be found at https://github.com/zhentingsheng/Note2Chat.

cs.CL↗
arXiv2026

RubricHub: A Comprehensive and Highly Discriminative Rubric Dataset via Automated Coarse-to-Fine Generation

Reinforcement Learning with Verifiable Rewards (RLVR) has driven substantial progress in reasoning-intensive domains like mathematics. However, optimizing open-ended generation remains challenging due to the lack of ground truth. While rubric-based evaluation offers a structured proxy for verification, existing methods suffer from scalability bottlenecks and coarse criteria, resulting in a supervision ceiling effect. To address this, we propose an automated Coarse-to-Fine Rubric Generation framework. By synergizing principle-guided synthesis, multi-model aggregation, and difficulty evolution, our approach produces comprehensive and highly discriminative criteria capable of capturing the subtle nuances. Based on this framework, we introduce RubricHub, a large-scale ($\sim$110k) and multi-domain dataset. We validate its utility through a two-stage post-training pipeline comprising Rubric-based Rejection Sampling Fine-Tuning (RuFT) and Reinforcement Learning (RuRL). Experimental results demonstrate that RubricHub unlocks significant performance gains: our post-trained Qwen3-14B achieves state-of-the-art (SOTA) results on HealthBench (69.3), surpassing proprietary frontier models such as GPT-5. Our code is available at \href{https://github.com/teqkilla/RubricHub}{ this URL}.

cs.AI↗
arXiv2026

SeRL: Self-Play Reinforcement Learning for Large Language Models with Limited Data

Recent advances have demonstrated the effectiveness of Reinforcement Learning (RL) in improving the reasoning capabilities of Large Language Models (LLMs). However, existing works inevitably rely on high-quality instructions and verifiable rewards for effective training, both of which are often difficult to obtain in specialized domains. In this paper, we propose Self-play Reinforcement Learning (SeRL) to bootstrap LLM training with limited initial data. Specifically, SeRL comprises two complementary modules: self-instruction and self-rewarding. The former module generates additional instructions based on the available data at each training step, employing robust online filtering strategies to ensure instruction quality, diversity, and difficulty. The latter module introduces a simple yet effective majority-voting mechanism to estimate response rewards for additional instructions, eliminating the need for external annotations. Finally, SeRL performs conventional RL based on the generated data, facilitating iterative self-play learning. Extensive experiments on various reasoning benchmarks and across different LLM backbones demonstrate that the proposed SeRL yields results superior to its counterparts and achieves performance on par with those obtained by high-quality data with verifiable rewards. Our code is available at https://github.com/wantbook-book/SeRL.

cs.CL↗
arXiv2026

Aligning Medical Conversational AI through Online Reinforcement Learning with Information-Theoretic Rewards

We present Information Gain Fine-Tuning (IGFT), a novel approach for training medical conversational AI to conduct effective patient interviews and generate comprehensive History of Present Illness (HPI) without requiring pre-collected human conversations. IGFT combines online Group Relative Policy Optimization (GRPO) with information-theoretic rewards, enabling models to learn from self-generated conversations with simulated patients. Unlike existing approaches that rely on expensive expert-annotated conversations or static datasets, our online RL framework allows models to discover effective questioning strategies through exploration. Our key innovation is an information gain reward function that tracks which clinical entities such as symptoms, temporal patterns, and medical history, are revealed during conversation. Each question's reward is computed based on its expected information gain combined with GPT-4o-mini quality assessments across dimensions including clinical relevance, patient engagement, and specificity. This hybrid approach ensures models learn to ask targeted, clinically appropriate questions that efficiently gather diagnostic information. We fine-tune two models using LoRA: Llama-3.1-8B-Instruct and DeepSeek-R1-Distill-Qwen-7B (a reasoning-optimized model). Training exclusively on Avey data containing concise HPIs, we evaluate generalization to MIMIC data with longer, more elaborate HPIs. DeepSeek-R1-Distill-Qwen-7B (IGFT) achieves F1 scores of 0.408 on Avey (10.9% improvement over base) and 0.289 on MIMIC (12.9% improvement), while Llama-3.1-8B-Instruct (IGFT) reaches 0.384 and 0.336 respectively. Both models outperform OpenAI's model on MIMIC and surpass medical domain-specific baselines like HuatuoGPT and UltraMedical, which were optimized for single-turn medical QA rather than multi-turn conversations.

cs.AI↗
arXiv2026

UCCL-EP: Portable Expert-Parallel Communication

Mixture-of-Experts (MoE) workloads rely on expert parallelism (EP) to achieve high GPU efficiency. State-of-the-art EP communication systems such as DeepEP demonstrate strong performance but exhibit poor portability across heterogeneous GPU and NIC platforms. The poor portability is rooted in architecture: GPU-initiated token-level RDMA communication requires tight vertical integration between GPUs and NICs, e.g., GPU writes to NIC driver/MMIO interfaces. We present UCCL-EP, a portable EP communication system that delivers DeepEP-level performance across heterogeneous GPU and NIC hardware. UCCL-EP replaces GPU-initiated RDMA with a high-throughput GPU-CPU control channel: compact token-routing commands are transferred to multithreaded CPU proxies, which then issue GPUDirect RDMA operations on behalf of GPUs. UCCL-EP further emulates various ordering semantics required by specialized EP communication modes using RDMA immediate data, enabling correctness on NICs that lack such ordering, e.g., AWS EFA. We implement UCCL-EP on NVIDIA and AMD GPUs with EFA and Broadcom NICs. On EFA, it outperforms the best existing EP solution by up to $2.1\times$ for dispatch and combine throughput. On NVIDIA-only platform, UCCL-EP achieves comparable performance to the original DeepEP. UCCL-EP also improves token throughput on SGLang by up to 40% on the NVIDIA+EFA platform, and improves DeepSeek-V3 training throughput over the AMD Primus/Megatron-LM framework by up to 45% on a 16-node AMD+Broadcom platform.

cs.DC↗
arXiv2026

Cyclic Modulation Control of Multi-Conflict Connected Automated Traffic

Multi-conflict traffic is ubiquitous. Connected Automated Vehicles (CAVs) offer unprecedented opportunities to enhance safety, reduce emissions, and increase throughput through precise coordination and automation. However, existing CAV strategies remain confined to specialized scenarios, such as highway on-ramp merging or single-lane roundabouts, and traditional traffic signals sacrifice efficiency for safety via rigid phasing and all-red intervals. In this paper, we present Cyclic Modulation Control of Multi-Conflict Connected Automated Traffic (CMAT), a unified, geometry-agnostic framework that embeds each conflict point into a repeating sequence of "micro-phases". Vehicles dynamically form platoons with demand-responsive sizes and negotiate time slots for occupying conflict points, enabling collision-free traversal and high intersection utilization. CMAT aims to minimize delay, guarantee safety, and accommodate arbitrary merging, diverging, and crossing patterns without manual retuning. We formalize CMAT as a mixed-integer linear programming model constructed on a directed graph abstracted from the physical intersection layout. The performance of CMAT is evaluated across a suite of multi-conflict tests, including simple two-way crossings, four-leg intersections, complex connected intersections. The results demonstrate substantial reductions in delay and significant throughput improvements compared with state-of-the-art CAV coordination methods and traditional signal timing strategies.

math.OC↗
arXiv2026

Reflected multi-entropy and its holographic dual

We introduce a mixed-state generalization of the multi-entropy through the canonical purification, which we call reflected multi-entropy. We propose the holographic dual of this measure. For the tripartite case, a field-theoretical calculation is performed using a six-point function of twist operators at large $c$ limit. At both zero and finite temperature, the field-theoretical results match the holographic results, supporting our holographic conjecture of this new measure.

hep-th↗
arXiv2026

CTIS-QA: Clinical Template-Informed Slide-level Question Answering for Pathology

In this paper, we introduce a clinical diagnosis template-based pipeline to systematically collect and structure pathological information. In collaboration with pathologists and guided by the the College of American Pathologists (CAP) Cancer Protocols, we design a Clinical Pathology Report Template (CPRT) that ensures comprehensive and standardized extraction of diagnostic elements from pathology reports. We validate the effectiveness of our pipeline on TCGA-BRCA. First, we extract pathological features from reports using CPRT. These features are then used to build CTIS-Align, a dataset of 80k slide-description pairs from 804 WSIs for vision-language alignment training, and CTIS-Bench, a rigorously curated VQA benchmark comprising 977 WSIs and 14,879 question-answer pairs. CTIS-Bench emphasizes clinically grounded, closed-ended questions (e.g., tumor grade, receptor status) that reflect real diagnostic workflows, minimize non-visual reasoning, and require genuine slide understanding. We further propose CTIS-QA, a Slide-level Question Answering model, featuring a dual-stream architecture that mimics pathologists' diagnostic approach. One stream captures global slide-level context via clustering-based feature aggregation, while the other focuses on salient local regions through attention-guided patch perception module. Extensive experiments on WSI-VQA, CTIS-Bench, and slide-level diagnostic tasks show that CTIS-QA consistently outperforms existing state-of-the-art models across multiple metrics. Code and data are available at https://github.com/HLSvois/CTIS-QA.

cs.CV↗
arXiv2026

NextFlow: Unified Sequential Modeling Activates Multimodal Understanding and Generation

We present NextFlow, a unified decoder-only autoregressive transformer trained on 6 trillion interleaved text-image discrete tokens. By leveraging a unified vision representation within a unified autoregressive architecture, NextFlow natively activates multimodal understanding and generation capabilities, unlocking abilities of image editing, interleaved content and video generation. Motivated by the distinct nature of modalities - where text is strictly sequential and images are inherently hierarchical - we retain next-token prediction for text but adopt next-scale prediction for visual generation. This departs from traditional raster-scan methods, enabling the generation of 1024x1024 images in just 5 seconds - orders of magnitude faster than comparable AR models. We address the instabilities of multi-scale generation through a robust training recipe. Furthermore, we introduce a prefix-tuning strategy for reinforcement learning. Experiments demonstrate that NextFlow achieves state-of-the-art performance among unified models and rivals specialized diffusion baselines in visual quality.

cs.CV↗
arXiv2026

Nonlinear Oscillatory Response of Automated Vehicle Car-following: Theoretical Analysis with Traffic State and Control Input Limits

This paper presents a framework grounded in the theory of describing function (DF) and incremental-input DF to theoretically analyze the nonlinear oscillatory response of automated vehicles (AVs) car-following (CF) amidst traffic oscillations, considering the limits of traffic state and control input. While prevailing approaches largely ignore these limits (i.e., saturation of acceleration/deceleration and speed) and focus on linear string stability analysis, this framework establishes a basis for theoretically analyzing the frequency response of AV systems with nonlinearities imposed by these limits. To this end, trajectories of CF pairs are decomposed into nominal and oscillatory trajectories, subsequently, the controlled AV system is repositioned within the oscillatory trajectory coordinates. Built on this base, DFs are employed to approximate the frequency responses of nonlinear saturation components by using their first harmonic output, thereby capturing the associated amplification ratio and phase shift. Considering the closed-loop nature of AV control systems, where system states and control input mutually influence each other, amplification ratios and phase shifts are balanced within the loop to ensure consistency. This balancing process may render multiple solutions, hence the incremental-input DF is further applied to identify the reasonable ones. The proposed method is validated by estimations from Simulink, and further comparisons with prevailing methods are conducted. Results confirm the alignment of our framework with Simulink results and exhibit its superior accuracy in analysis compared to the prevailing methods. Furthermore, the framework proves valuable in string stability analysis, especially when conventional linear methods offer misleading insights.

eess.SY↗
arXiv2026

Explainability-Guided Defense: Attribution-Aware Model Refinement Against Adversarial Data Attacks

The growing reliance on deep learning models in safety-critical domains such as healthcare and autonomous navigation underscores the need for defenses that are both robust to adversarial perturbations and transparent in their decision-making. In this paper, we identify a connection between interpretability and robustness that can be directly leveraged during training. Specifically, we observe that spurious, unstable, or semantically irrelevant features identified through Local Interpretable Model-Agnostic Explanations (LIME) contribute disproportionately to adversarial vulnerability. Building on this insight, we introduce an attribution-guided refinement framework that transforms LIME from a passive diagnostic into an active training signal. Our method systematically suppresses spurious features using feature masking, sensitivity-aware regularization, and adversarial augmentation in a closed-loop refinement pipeline. This approach does not require additional datasets or model architectures and integrates seamlessly into standard adversarial training. Theoretically, we derive an attribution-aware lower bound on adversarial distortion that formalizes the link between explanation alignment and robustness. Empirical evaluations on CIFAR-10, CIFAR-10-C, and CIFAR-100 demonstrate substantial improvements in adversarial robustness and out-of-distribution generalization.

cs.LG↗
arXiv2026

AutoTrust: Benchmarking Trustworthiness in Large Vision Language Models for Autonomous Driving

Recent advancements in large vision language models (VLMs) tailored for autonomous driving (AD) have shown strong scene understanding and reasoning capabilities, making them undeniable candidates for end-to-end driving systems. However, limited work exists on studying the trustworthiness of DriveVLMs -- a critical factor that directly impacts public transportation safety. In this paper, we introduce AutoTrust, a comprehensive trustworthiness benchmark for large vision-language models in autonomous driving (DriveVLMs), considering diverse perspectives -- including trustfulness, safety, robustness, privacy, and fairness. We constructed the largest visual question-answering dataset for investigating trustworthiness issues in driving scenarios, comprising over 10k unique scenes and 18k queries. We evaluated six publicly available VLMs, spanning from generalist to specialist, from open-source to commercial models. Our exhaustive evaluations have unveiled previously undiscovered vulnerabilities of DriveVLMs to trustworthiness threats. Specifically, we found that the general VLMs like LLaVA-v1.6 and GPT-4o-mini surprisingly outperform specialized models fine-tuned for driving in terms of overall trustworthiness. DriveVLMs like DriveLM-Agent are particularly vulnerable to disclosing sensitive information. Additionally, both generalist and specialist VLMs remain susceptible to adversarial attacks and struggle to ensure unbiased decision-making across diverse environments and populations. Our findings call for immediate and decisive action to address the trustworthiness of DriveVLMs -- an issue of critical importance to public safety and the welfare of all citizens relying on autonomous transportation systems. We release all the codes and datasets in https://github.com/taco-group/AutoTrust.

cs.CV↗