BZPEERReading, trust and collaboration for research
Menu

Discover

GraphFollow connections around a paper, topic or saved view.

Workspaces

HomePick up where your research flow left off.InboxHandle requests, replies and notifications from one place.CollectionsCurate reading lists, evidence sets and shareable dossiers.ResearchSee your private provenance graph, trail and imports.CollaborationMove from discovery into focused discussion rooms.PublishDraft, preview and publish full-text research articles.

Network

ExpertsFind specialists worth following or asking for help.CommunitiesJoin research circles organized around shared work.PartnersSee external organizations active around the same topics.

Opportunities

ListingsBrowse roles, calls and collaborations with clearer fit signals.

Account

Log inReturn to your reading and collaboration workspace.Create accountStart saving work, following people and joining teams.
Log inCreate account

Researcher profile

Ziyue Wang

Ziyue Wang contributes to research discovery and scholarly infrastructure.

ResearcherAffiliation not importedOpen to collaborate
Machine LearningArtificial IntelligenceComputer Visionhep-phMultimediaastro-ph.HEastro-ph.IMComputation and Languageeess.IVgr-qchep-thmath.STnucl-thphysics.data-anphysics.soc-phStatistics Theory

Trust snapshot

Quick read

Trust 21 - EmergingVerification L1Unclaimed author
15works
0followers
16topics
4close collaborators

Actions

Decide how to stay connected

Follow researcher0

Identity and collaboration

How to connect with this researcher

Claiming links this public author record to a researcher profile and unlocks direct collaboration workflows.

Log in to claim

Direct collaboration

Open a focused conversation when the fit is right

Claim this author entity first to unlock direct invitations.

Research graph

See the researcher in context

Open full explorer

Inspect adjacent work, topics, institutions and collaborators without jumping out to a separate graph page.

Building this graph slice

BZPEER is loading the nearby papers, people, topics and institutions for this page.

Published work

15 published item(s)

preprint2026arXiv

Browse and Concentrate: Comprehending Multimodal Content via prior-LLM Context Fusion

With the bloom of Large Language Models (LLMs), Multimodal Large Language Models (MLLMs) that incorporate LLMs with pre-trained vision models have recently demonstrated impressive performance across diverse vision-language tasks. However, they fall short to comprehend context involving multiple images. A primary reason for this shortcoming is that the visual features for each images are encoded individually by frozen encoders before feeding into the LLM backbone, lacking awareness of other images and the multimodal instructions. We term this issue as prior-LLM modality isolation and propose a two phase paradigm, browse-and-concentrate, to enable in-depth multimodal context fusion prior to feeding the features into LLMs. This paradigm initially "browses" through the inputs for essential insights, and then revisits the inputs to "concentrate" on crucial details, guided by these insights, to achieve a more comprehensive understanding of the multimodal inputs. Additionally, we develop training strategies specifically to enhance the understanding of multi-image inputs. Our method markedly boosts the performance on 7 multi-image scenarios, contributing to increments on average accuracy by 2.13% and 7.60% against strong MLLMs baselines with 3B and 11B LLMs, respectively.

0 citations0 reviewsNo code linkedOpen access
preprint2026arXiv

Distributional Spectral Diagnostics for Localizing Grokking Transitions

In grokking, a model first fits the training data while test accuracy remains low, and only later begins to generalize. We ask whether this transition can be localized from observed training trajectories before the test accuracy rises, and formulate grokking transition localization as a diagnostic problem with an explicit threshold/FPR/lead-time trade-off. Task-dependent observables are summarized as empirical distributions, mapped to Wasserstein/quantile coordinates, and analyzed by Hankel dynamic mode decomposition (DMD); the resulting reconstruction residual, together with spectrum and effective rank, forms the diagnostic output. On held-out modular-addition Transformer runs, the residual achieves AUROC \(\approx \) 0.93 for grokking-vs-non-grokking discrimination at the run level; under a fixed sustained-threshold operating rule, true-positive alarms can precede onset, with lead time reported jointly with false-alarm rate and uncertainty intervals. Perturbation experiments show that, in the tested \(wd=1\) pool, high-residual windows exhibit about \(3\times\) larger short-horizon perturbation deviation than low-residual windows. In a same-data norm-window control, perturbation sensitivity aligns with the residual ordering rather than total-parameter-norm ordering, suggesting that the residual is not merely a total-norm proxy at the window level in the studied \(wd=1\) dynamics. Norm signals remain strong run-level regime indicators, and log-probability performs best among the observables tested under the current protocol. We position the residual as a window-level monitoring and localization signal in the studied modular-arithmetic Transformer settings, not a universal early-warning predictor or an intervention rule.

0 citations0 reviewsNo code linkedOpen access
preprint2026arXiv

Leveraging Error Diversity in Group Rollouts for Reinforcement Learning

Reinforcement Learning from Verifiable Rewards (RLVR) typically samples multiple responses per prompt and assigns binary rewards based on individual correctness, yet the collective structure of the group output, specifically the distribution of errors, is largely discarded. We identify this as a missed opportunity: empirical analysis reveals that error diversity within a group is a strong predictor of training success, with problems eliciting diverse wrong answers benefiting substantially more from RLVR than those producing homogeneous failures. Motivated by this observation, we propose Error Diversity Advantage Shaping (EDAS), a lightweight, algorithm-agnostic technique that modulates the advantage signal for incorrect rollouts based on intra-group error diversity. EDAS amplifies penalties for dominant, repeated errors and attenuates penalties for rare, exploratory ones, thereby encouraging the model to maintain diverse reasoning paths and discouraging error perseveration. Crucially, EDAS operates as a simple post-hoc adjustment that can be seamlessly integrated into any RLVR algorithm. We validate EDAS on top of several mainstream RLVR methods across a series of models and seven challenging math benchmarks, demonstrating consistent improvements. Notably, EDAS yields an average improvement of 6.29 points over DAPO on Qwen3-8B across seven benchmarks, confirming that exploiting the latent information in group rollouts is a broadly effective strategy for strengthening RLVR.

0 citations0 reviewsNo code linkedOpen access
preprint2026arXiv

Patient-Zero: Scaling Synthetic Patient Agents to Real-World Distributions without Real Patient Data

Synthetic data generation with Large Language Models (LLMs) has emerged as a promising solution in the medical domain to mitigate data scarcity and privacy constraints. However, existing approaches remain constrained by their derivative nature, relying on real-world records, which pose privacy risks and distribution biases. Furthermore, current patient agents face the Stability-Plasticity Dilemma, struggling to maintain clinical consistency during dynamic inquiries. To address these challenges, we introduce Patient-Zero, a novel framework for ab initio patient simulation that requires no real medical records. Our Medically-Aligned Hierarchical Synthesis framework generates comprehensive and diverse patient records from abstract clinical guidelines via stratified attribute permutation. To support rigorous clinical interaction, we design a Dual-Track Cognitive Memory System to enable agents dynamically update memory while preserving logical consistency and persona adherence. Extensive evaluations show that Patient-Zero establishes a new state-of-the-art in both data quality and interaction fidelity. In human expert evaluations, senior licensed physicians judge our synthetic data to be statistically indistinguishable from real human-authored data and higher in clinical quality. Furthermore, downstream medical reasoning model trained on our synthetic dataset shows substantial performance gains (MedQA +24.0%; MMLU +14.5%), demonstrating the practical utility of our framework.

0 citations0 reviewsNo code linkedOpen access
preprint2026arXiv

Step-wise Rubric Rewards for LLM Reasoning

Reinforcement Learning with Verifiable Rewards (RLVR) is widely used to improve reasoning in large language models, but rewards only final-answer correctness with no supervision over intermediate steps. Rubric-based methods such as Rubrics as Rewards (RaR) introduce finer-grained supervision by scoring rollouts against structured criteria, yet the rubric scores are still aggregated into a single scalar applied to the entire response, causing three weaknesses: loss of multi-criterion structure, uniform supervision of correct and incorrect steps, and reward hacking through unbounded self-correction. On 1,000 problems, we find 18.2% of steps in correct-answer responses are wrong yet positively rewarded, while 49.9% of steps in incorrect-answer responses are correct yet penalized. We introduce Step-wise Rubrics as Rewards (SRaR), an RLVR framework that (i) uses an LLM judge to attribute each rubric item to a specific reasoning step, (ii) normalizes per-step rubric scores across rollouts so only steps whose quality varies produce a learning signal, and (iii) combines the per-step reward with the outcome reward through a decoupled advantage estimator that keeps the outcome baseline stable. We further build a 16K-problem rubric dataset by contrastively distilling rubric items from correct and flawed reasoning paths sampled from a strong model. Across six mathematical reasoning benchmarks, SRaR improves average accuracy over RaR by 3.57 points on Qwen3-8B and 2.75 points on Qwen3-32B, raises the Faithful Reasoning Rate on AIME 2025 from 34.5% to 46.7%, and reduces self-correction looping from 48.1% to 26.5%.

0 citations0 reviewsNo code linkedOpen access
preprint2026arXiv

When Agents Overtrust Environmental Evidence: An Extensible Agentic Framework for Benchmarking Evidence-Grounding Defects in LLM Agents

Large language model agents increasingly operate through environment-facing scaffolds that expose files, web pages, APIs, and logs. These observations influence tool use, state tracking, and action sequencing, yet their reliability and authority are often uncertain. Environmental grounding is therefore a systems-level problem involving context admission, evidence provenance, freshness checking, verification policy, action gating, and model reasoning. Existing agent benchmarks mainly evaluate task capability or specific attacks such as prompt injection and memory poisoning, but they under-specify a fundamental reliability question: whether agents remain grounded in the true environment state when observations are stale, incorrect, or malicious. We introduce EnvTrustBench, an agentic framework for benchmarking this failure mode. We define an evidence-grounding defect (EGD) as a behavioral failure in which an agent treats an environment-facing claim as sufficient evidence for action without resolving it against available current evidence, leading to a task-incorrect false path under the true environment state. Given a task scenario, EnvTrustBench generates the workspace, environment, agent-facing objective, and validation oracle, executes the evaluated agent, records its action-observation trajectory and final state, and applies the oracle to produce a verdict. Using 6 LLM backbones and 5 widely used scaffolds, we evaluate 55 generated cases across 11 task scenarios, with each scenario expanded through five feedback-guided generation iterations. Results show that EGDs consistently emerge across operational workflows, highlighting environmental grounding as a core agent reliability problem with important security implications.

0 citations0 reviewsNo code linkedOpen access
preprint2025arXiv

Learning to detect continuous gravitational waves: an open data-analysis competition

We report results of a public data-analysis challenge, hosted on the open data-science platform Kaggle, to detect simulated continuous gravitational-wave signals (CWs). These are weak signals from rapidly spinning neutron stars that remain undetected despite extensive searches. The competition dataset consisted of a population of CW signals using both simulated and real LIGO detector data matching the conditions of actual CW searches. The competition attracted more than 1,000 participants to develop realistic CW search algorithms. We describe the top 10 approaches and discuss their applicability as a pre-processing step compared to standard CW-search approaches. For the competition's dataset, we find that top approaches can reduce the computing cost by 1 to 3 orders of magnitude at a false-dismissal probability comparable to standard CW searches. Additionally, the competition drove the development of new GPU-accelerated detection pipelines, which facilitated their adoption in other areas of gravitational-wave data analysis. We release the associated dataset, which constitutes the first open standardized benchmark for CW detection, to enable reproducible method comparisons and to encourage further developments toward the first detection of these elusive signals.

0 citations0 reviewsNo code linkedOpen access
preprint2023arXiv

Imperceptible Adversarial Attack via Invertible Neural Networks

Adding perturbations via utilizing auxiliary gradient information or discarding existing details of the benign images are two common approaches for generating adversarial examples. Though visual imperceptibility is the desired property of adversarial examples, conventional adversarial attacks still generate traceable adversarial perturbations. In this paper, we introduce a novel Adversarial Attack via Invertible Neural Networks (AdvINN) method to produce robust and imperceptible adversarial examples. Specifically, AdvINN fully takes advantage of the information preservation property of Invertible Neural Networks and thereby generates adversarial examples by simultaneously adding class-specific semantic information of the target class and dropping discriminant information of the original class. Extensive experiments on CIFAR-10, CIFAR-100, and ImageNet-1K demonstrate that the proposed AdvINN method can produce less imperceptible adversarial images than the state-of-the-art methods and AdvINN yields more robust adversarial examples with high confidence compared to other adversarial attacks.

0 citations0 reviewsNo code linkedOpen access
preprint2022arXiv

Learn to Understand Negation in Video Retrieval

Negation is a common linguistic skill that allows human to express what we do NOT want. Naturally, one might expect video retrieval to support natural-language queries with negation, e.g., finding shots of kids sitting on the floor and not playing with a dog. However, the state-of-the-art deep learning based video retrieval models lack such ability, as they are typically trained on video description datasets such as MSR-VTT and VATEX that lack negated descriptions. Their retrieved results basically ignore the negator in the sample query, incorrectly returning videos showing kids playing with dog. This paper presents the first study on learning to understand negation in video retrieval and make contributions as follows. By re-purposing two existing datasets (MSR-VTT and VATEX), we propose a new evaluation protocol for video retrieval with negation. We propose a learning based method for training a negation-aware video retrieval model. The key idea is to first construct a soft negative caption for a specific training video by partially negating its original caption, and then compute a bidirectionally constrained loss on the triplet. This auxiliary loss is weightedly added to a standard retrieval loss. Experiments on the re-purposed benchmarks show that re-training the CLIP (Contrastive Language-Image Pre-Training) model by the proposed method clearly improves its ability to handle queries with negation. In addition, the model performance on the original benchmarks is also improved.

0 citations0 reviewsNo code linkedOpen access
preprint2022arXiv

Lightweight Attentional Feature Fusion: A New Baseline for Text-to-Video Retrieval

In this paper we revisit feature fusion, an old-fashioned topic, in the new context of text-to-video retrieval. Different from previous research that considers feature fusion only at one end, let it be video or text, we aim for feature fusion for both ends within a unified framework. We hypothesize that optimizing the convex combination of the features is preferred to modeling their correlations by computationally heavy multi-head self attention. We propose Lightweight Attentional Feature Fusion (LAFF). LAFF performs feature fusion at both early and late stages and at both video and text ends, making it a powerful method for exploiting diverse (off-the-shelf) features. The interpretability of LAFF can be used for feature selection. Extensive experiments on five public benchmark sets (MSR-VTT, MSVD, TGIF, VATEX and TRECVID AVS 2016-2020) justify LAFF as a new baseline for text-to-video retrieval.

0 citations0 reviewsNo code linkedOpen access
preprint2022arXiv

Tractable and Near-Optimal Adversarial Algorithms for Robust Estimation in Contaminated Gaussian Models

Consider the problem of simultaneous estimation of location and variance matrix under Huber's contaminated Gaussian model. First, we study minimum $f$-divergence estimation at the population level, corresponding to a generative adversarial method with a nonparametric discriminator and establish conditions on $f$-divergences which lead to robust estimation, similarly to robustness of minimum distance estimation. More importantly, we develop tractable adversarial algorithms with simple spline discriminators, which can be implemented via nested optimization such that the discriminator parameters can be fully updated by maximizing a concave objective function given the current generator. The proposed methods are shown to achieve minimax optimal rates or near-optimal rates depending on the $f$-divergence and the penalty used. This is the first time such near-optimal error rates are established for adversarial algorithms with linear discriminators under Huber's contamination model. We present simulation studies to demonstrate advantages of the proposed methods over classic robust estimators, pairwise methods, and a generative adversarial method with neural network discriminators.

0 citations0 reviewsNo code linkedOpen access
preprint2021arXiv

Spin Polarization Induced by Inhomogeneous Dynamical Condensate

The role of dynamical chiral condensate in spin polarization is investigated in a kinetic theory framework. Transport equations for quark matter are derived in the mean-field approximation for the Nambu--Jona-Lasinio model. The dynamical condensate carries part of the energy momentum tensor (EMT) and the angular momentum tensor (AMT), the conservation of EMT and AMT can be proved from the kinetic equations as required by the symmetry. The transport equations of vector and axial-vector components are derived taking the spin decomposition as well as semi-classical expansion. Inhomogeneous mass introduces novel effect at $\mathcal{O}(\hbar)$, for an initially unpolarized system, spin polarization can be generated from the dynamical chiral condensate, even without the collision term. The stable spin distribution function is found to be robust, namely in the case with non-trivial dynamical mass, the spin polarization is still enslaved by the thermal vorticity, while the Killing condition can be loosened.

0 citations0 reviewsNo code linkedOpen access
preprint2020arXiv

Chiral Phase Transition in an Expanding Quark System

We investigate the influence of chiral symmetry which varies along the space-time evolution of the system by considering the chiral phase transition in an non-equilibrium expanding quark-antiquark system. The chiral symmetry is described by the mean field order parameter, whose values is the solution of a self-consistent equation, and affects the space-time evolution of the system through the force term in the Vlasov equation. The Vlasov equation and the gap equation are solved concurrently and continuously for a longitudinal boost-invariant and transversely rotation-invariant system. This numerical framework enables us to carefully investigate how the phase transition and collision affect the evolution of the system. It is observed that the chiral phase transition gives rise to a kink in the flow velocity, which is caused by the force term in the Vlasov equation. The kink is enhanced by larger susceptibility and tends to be smoothed out by non-equilibrium effect. The spatial phase boundary appears as a "wall" for the quarks, as the quarks with low momentum are bounced back, while those with high momentum go through the wall but are slowed down.

0 citations0 reviewsNo code linkedOpen access
preprint2020arXiv

Entity Profiling in Knowledge Graphs

Knowledge Graphs (KGs) are graph-structured knowledge bases storing factual information about real-world entities. Understanding the uniqueness of each entity is crucial to the analyzing, sharing, and reusing of KGs. Traditional profiling technologies encompass a vast array of methods to find distinctive features in various applications, which can help to differentiate entities in the process of human understanding of KGs. In this work, we present a novel profiling approach to identify distinctive entity features. The distinctiveness of features is carefully measured by a HAS model, which is a scalable representation learning model to produce a multi-pattern entity embedding. We fully evaluate the quality of entity profiles generated from real KGs. The results show that our approach facilitates human understanding of entities in KGs.

0 citations0 reviewsNo code linkedOpen access
preprint2020arXiv

Mass Correction to Chiral Kinetic Equations

We study fermion mass correction to chiral kinetic equations in electromagnetic fields. Different from the chiral limit where fermion number density is the only independent distribution, the number and spin densities are coupled to each other for massive fermion systems. To the first order in $\hbar$, we derived the quantum correction to the classical on-shell condition and the Boltzmann-type transport equations. To the linear order in the fermion mass, the mass correction does not change the structure of the chiral kinetic equations and behaves like additional collision terms. While the mass correction exists already at classical level in general electromagnetic fields, it is only a first order quantum correction in the study of chiral magnetic effect.

0 citations0 reviewsNo code linkedOpen access