BZPEERPapers, people, evidence and research paths
Menu

Discover

SearchFind papers, people, topics, institutions and opportunities from one place.GraphSee how papers, authors, topics and evidence connect.

Research tools

HomeContinue from the papers, people and topics that matter now.CollectionsTurn saved papers into reading lists, evidence maps and shareable context.ResearchReconstruct the path that led you from search to paper to question.CollaborationMove from a paper, topic or expert request into a focused thread.PublishCreate full-text research articles with metadata, formulas and versions.

Network

ExpertsFind specialists whose work and reviews explain why they can help.CommunitiesJoin research groups organized around papers, topics and shared evidence.

Opportunities

ListingsBrowse roles, calls and collaborations connected to real research context.

Account

Log inReturn to your saved papers, graph views and active threads.Create accountStart saving papers, building a research trail and joining teams.
Log inCreate account

Source author record

Ye Lu

Ye Lu appears in the imported research catalog. Authorship, coauthor and topic links are available while profile ownership is still unclaimed.

ResearcherUnclaimed source record
astro-ph.HEcond-mat.mes-hallcond-mat.mtrl-scimath.NANumerical AnalysisArtificial Intelligenceastro-ph.COastro-ph.GAComputational Engineering, Finance, and ScienceComputer Visioncs.CYMachine Learningphysics.data-an

Catalog footprint

What is connected

10works
13topics
4close collaborators

Actions

Connect this record

Log in to claim
Open graphBrowse works

Research graph

See the researcher in context

Open full explorer

Inspect adjacent papers, topics, institutions and collaborators without losing the researcher page.

Building this map preview

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

Published work

10 published item(s)

preprint2026arXiv

AcademiClaw: When Students Set Challenges for AI Agents

Benchmarks within the OpenClaw ecosystem have thus far evaluated exclusively assistant-level tasks, leaving the academic-level capabilities of OpenClaw largely unexamined. We introduce AcademiClaw, a bilingual benchmark of 80 complex, long-horizon tasks sourced directly from university students' real academic workflows -- homework, research projects, competitions, and personal projects -- that they found current AI agents unable to solve effectively. Curated from 230 student-submitted candidates through rigorous expert review, the final task set spans 25+ professional domains, ranging from olympiad-level mathematics and linguistics problems to GPU-intensive reinforcement learning and full-stack system debugging, with 16 tasks requiring CUDA GPU execution. Each task executes in an isolated Docker sandbox and is scored on task completion by multi-dimensional rubrics combining six complementary techniques, with an independent five-category safety audit providing additional behavioral analysis. Experiments on six frontier models show that even the best achieves only a 55\% pass rate. Further analysis uncovers sharp capability boundaries across task domains, divergent behavioral strategies among models, and a disconnect between token consumption and output quality, providing fine-grained diagnostic signals beyond what aggregate metrics reveal. We hope that AcademiClaw and its open-sourced data and code can serve as a useful resource for the OpenClaw community, driving progress toward agents that are more capable and versatile across the full breadth of real-world academic demands. All data and code are available at https://github.com/GAIR-NLP/AcademiClaw.

0 citations0 reviewsNo code linkedOpen access
preprint2022arXiv

Stochastic additive manufacturing simulation: from experiment to surface roughness and porosity prediction

Deterministic computational modeling of laser powder bed fusion (LPBF) process fails to capture irregularities and roughness of the scan track, unless expensive powder-scale analysis is used. In this work we developed a stochastic computational modeling framework based on Markov Chain Monte Carlo (MCMC) capable of capturing the irregularities of LPBF scan. The model is calibrated against AFRL single track scan data using a specially designed tensor decomposition method, i.e., Higher-Order Proper Generalized Decomposition (HOPGD) that relies on non-intrusive data learning and construction of reduced order surrogate models. Once calibrated, the stochastic model can be used to predict the roughness and porosity at part scale at a significantly reduced computational cost compared to detailed powder-scale deterministic simulations. The stochastic simulation predictions are validated against AFRL multi-layer and multitrack experiments and reported as more accurate when compared with regular deterministic simulation results.

0 citations0 reviewsNo code linkedOpen access
preprint2021arXiv

HiDeNN-PGD: reduced-order hierarchical deep learning neural networks

This paper presents a proper generalized decomposition (PGD) based reduced-order model of hierarchical deep-learning neural networks (HiDeNN). The proposed HiDeNN-PGD method keeps both advantages of HiDeNN and PGD methods. The automatic mesh adaptivity makes the HiDeNN-PGD more accurate than the finite element method (FEM) and conventional PGD, using a fraction of the FEM degrees of freedom. The accuracy and convergence of the method have been studied theoretically and numerically, with a comparison to different methods, including FEM, PGD, HiDeNN and Deep Neural Networks. In addition, we theoretically showed that the PGD converges to FEM at increasing modes, and the PGD error is a direct sum of the FEM error and the mode reduction error. The proposed HiDeNN-PGD performs high accuracy with orders of magnitude fewer degrees of freedom, which shows a high potential to achieve fast computations with a high level of accuracy for large-size engineering problems.

0 citations0 reviewsNo code linkedOpen access
preprint2020arXiv

VecQ: Minimal Loss DNN Model Compression With Vectorized Weight Quantization

Quantization has been proven to be an effective method for reducing the computing and/or storage cost of DNNs. However, the trade-off between the quantization bitwidth and final accuracy is complex and non-convex, which makes it difficult to be optimized directly. Minimizing direct quantization loss (DQL) of the coefficient data is an effective local optimization method, but previous works often neglect the accurate control of the DQL, resulting in a higher loss of the final DNN model accuracy. In this paper, we propose a novel metric called Vector Loss. Based on this new metric, we develop a new quantization solution called VecQ, which can guarantee minimal direct quantization loss and better model accuracy. In addition, in order to speed up the proposed quantization process during model training, we accelerate the quantization process with a parameterized probability estimation method and template-based derivation calculation. We evaluate our proposed algorithm on MNIST, CIFAR, ImageNet, IMDB movie review and THUCNews text data sets with numerical DNN models. The results demonstrate that our proposed quantization solution is more accurate and effective than the state-of-the-art approaches yet with more flexible bitwidth support. Moreover, the evaluation of our quantized models on Saliency Object Detection (SOD) tasks maintains comparable feature extraction quality with up to 16$\times$ weight size reduction.

0 citations0 reviewsNo code linkedOpen access
preprint2014arXiv

Relativistic Jet Properties of GeV-TeV Blazars and Possible Implications for the Jet Formation, Composition, and Cavity Kinematics

We fit the spectral energy distributions (SEDs) of a GeV-TeV FSRQ sample with the leptonic model. Their gamma_min of the relativistic electron distributions, which significantly affect the estimates of the jet properties, are constrained, with a typical value of 48. Their jet power, magnetized parameter, radiation efficiency, and jet production/radiation rates per central black hole (BH) mass are derived and compared to that of BL Lacs. We show that he FSRQ jets may be dominated by the Poynting flux and have a high radiation efficiency, whereas the BL Lac jets are likely dominated by particles and have a lower radiation efficiency than FSRQs. Being different from BL Lacs, the jet powers of FSRQs are proportional to their central BH masses. The jet production and radiation rates of the FSRQs distribute in narrow ranges and are correlated with each other, whereas no similar feature is found for the BL Lacs. We also show that the jet power is correlated with the cavity kinetic power, and the magnetic field energy in the jets may provide the cavity kinetic energy of FSRQs and the kinetic energy of cold protons in the jets may be crucial for cavity kinetic energy of BL Lacs. We suggest that the dominating formation mechanism of FSRQ jets may be the BZ process, but BL Lac jets may be produced via the BP and/or BZ processes, depending on the structures and accretion rates of accretion disks.

0 citations0 reviewsNo code linkedOpen access
preprint2013arXiv

Radiation Mechanism and Jet Composition of Gamma-Ray Bursts and GeV-TeV selected Radio Loud Active Galactic Nuclei

Gamma-ray bursts (GRBs) and GeV-TeV selected radio loud Active Galactic Nuclei (AGNs) are compared based on our systematic modeling of the observed spectral energy distributions of a sample of AGNs with a single-zone leptonic model. We show that the correlation between the jet power (P_{jet}) and the prompt gamma-ray luminosity (L_{jet}) of GRBs is consistent, within the uncertainties, with the correlation between jet power and the synchrotron peak luminosity (L_{s, jet}) of flat spectrum radio quasars (FSRQs). Their radiation efficiencies (\varepsilon) are also comparable (>10% for most sources), which increase with the bolometric jet luminosity (L_{bol,jet}) for FSRQs and with the L_{jet} for GRBs with similar power-law indices. BL Lacs do not follow the P_{jet}-L_{s, jet} relation of FSRQs. They have lower \varepsilon and L_{bol, jet} values than FSRQs, and a tentative L_{bol, jet}-\varepsilon relation is also found, with a power-law index being different from that of the FSRQs. The magnetization parameters (σ) of FSRQs are averagely larger than that of BL Lacs. They are anti-correlated with $\varepsilon$ for the FSRQs, but positive correlated with \varepsilon for the BL Lacs. GeV Narrow-line Seyfert 1 galaxies potentially share similar properties with FSRQs. Based on the analogy between GRBs and FSRQs, we suggest that the prompt gamma-ray emission of GRBs is likely produced by synchrotron process in a magnetized jet with high radiation efficiency, similar to FSRQs. The jets of BL Lacs, on the other hand, are less efficient and are likely more matter dominated.

0 citations0 reviewsNo code linkedOpen access
preprint2013arXiv

Radiative Efficiency of Disk Accretion in Individual SDSS QSOs

We estimate the radiative efficiency (epsilon) of individual type 1 SDSS QSOs by using their bolometric luminosities (L_{bol}) and accretion rates (\dot{M}_{\bullet,acc}), which may be related to the assembly histories and spins of the central massive black holes (MBHs). We estimate L_{bol} by using the empirical spectral energy distributions of QSOs and \dot{M}_{\bullet,acc} by fitting the observed optical luminosity(/-ies) with the thin accretion disk model, assuming the MBH masses given by the virial mass estimator(s) (M_{\bullet,vir}). We find an apparent correlation between epsilon and M_{\bullet,vir}, which is strong at redshift z \la 1.8, weak at z \ga 2, and consistent with that found by Davis & Laor (2011) for 80 PG QSOs at z \leq 0.5. To investigate whether this correlation is intrinsic or not, we construct a mock sample of QSOs according to the true MBH mass and Eddington ratio distributions given in Kelly & Shen (2013). By comparing the results obtained from the mock sample with that from the SDSS sample, we demonstrate that the apparent epsilon-M_{\bullet,vir} correlation can be produced by and mainly due to the selection effects of the SDSS sample and the bias induced by the usage of M_{\bullet,vir} as the true MBH mass. The mean values of epsilon of those SDSS QSOs are consistent with being a constant \simeq 0.11-0.16 over the redshift range of 0.3 \la z \la 4. We conclude that the current SDSS QSO data is consistent with no strong intrinsic correlation between radiative efficiency and true MBH mass and no significant redshift evolution of radiative efficiencies.

0 citations0 reviewsNo code linkedOpen access
preprint2011arXiv

Graphene-protein bioelectronic devices with wavelength-dependent photoresponse

We implemented a nanoelectronic interface between graphene field effect transistors (FETs) and soluble proteins. This enables production of bioelectronic devices that combine functionalities of the biomolecular and inorganic components. The method serves to link polyhistidine-tagged proteins to graphene FETs using the tag itself. Atomic Force Microscopy and Raman spectroscopy provide structural understanding of the bio/nano hybrid; current-gate voltage measurements are used to elucidate the electronic properties. As an example application, we functionalize graphene FETs with fluorescent proteins to yield hybrids that respond to light at wavelengths defined by the optical absorption spectrum of the protein

0 citations0 reviewsNo code linkedOpen access
preprint2011arXiv

In-situ electronic characterization of graphene nanoconstrictions fabricated in a transmission electron microscope

We report electronic measurements on high-quality graphene nanoconstrictions (GNCs) fabricated in a transmission electron microscope (TEM), and the first measurements on GNC conductance with an accurate measurement of constriction width down to 1 nm. To create the GNCs, freely-suspended graphene ribbons were fabricated using few-layer graphene grown by chemical vapor deposition. The ribbons were loaded into the TEM, and a current-annealing procedure was used to clean the material and improve its electronic characteristics. The TEM beam was then used to sculpt GNCs to a series of desired widths in the range 1 - 700 nm; after each sculpting step, the sample was imaged by TEM and its electronic properties measured in-situ. GNC conductance was found to be remarkably high, comparable to that of exfoliated graphene samples of similar size. The GNC conductance varied with width approximately as, where w is the constriction width in nanometers. GNCs support current densities greater than 120 \muA/nm2, two orders of magnitude higher than has been previously reported for graphene nanoribbons and 2000 times higher than copper.

0 citations0 reviewsNo code linkedOpen access
preprint2010arXiv

DNA-decorated graphene chemical sensors

Graphene is a true two dimensional material with exceptional electronic properties and enormous potential for practical applications. Graphene's promise as a chemical sensor material has been noted but there has been relatively little work on practical chemical sensing using graphene, and in particular how chemical functionalization may be used to sensitize graphene to chemical vapors. Here we show one route towards improving the ability of graphene to work as a chemical sensor by using single stranded DNA as a sensitizing agent. The resulting broad response devices show fast response times, complete and rapid recovery to baseline at room temperature, and discrimination between several similar vapor analytes.

0 citations0 reviewsNo code linkedOpen access