arXiv subjects

Yawen Liu

Publications and source records attributed to Yawen Liu.

18 recordsLinked to original sources

arXiv2025

Millimeter Wave Inverse Pinhole Imaging

Millimeter wave (mmWave) radars are popular for perception in vision-denied contexts due to their compact size. This paper explores emerging use-cases that involve static mount or momentarily-static compact radars, for example, a hovering drone. The key challenge with static compact radars is that their limited form-factor also limits their angular resolution. This paper presents Umbra, a mmWave high resolution imaging system, that introduces the concept of rotating mmWave "inverse pinholes" for angular resolution enhancement. We present the imaging system model, design, and evaluation of mmWave inverse pinholes. The inverse pinhole is attractive for its lightweight nature, which enables low-power rotation, upgrading static-mount radars. We also show how propellers in aerial vehicles act as natural inverse pinholes and can enjoy the benefits of high-resolution imaging even while they are momentarily static, e.g., hovering. Our evaluation shows Umbra resolving up to 2.5$^{\circ}$ with just a single antenna, a 5$\times$ improvement compared to 14$^{\circ}$ from a compact mmWave radar baseline.

eess.IV↗

arXiv2025

Compass-Thinker-7B Technical Report

Recent R1-Zero-like research further demonstrates that reasoning extension has given large language models (LLMs) unprecedented reasoning capabilities, and Reinforcement Learning is the core technology to elicit its complex reasoning. However, conducting RL experiments directly on hyperscale models involves high computational costs and resource demands, posing significant risks. We propose the Compass-Thinker-7B model, which aims to explore the potential of Reinforcement Learning with less computational resources and costs, and provides insights for further research into RL recipes for larger models. Compass-Thinker-7B is trained from an open source model through a specially designed Reinforcement Learning Pipeline. We curate a dataset of 30k verifiable mathematics problems for the Reinforcement Learning Pipeline. By configuring data and training settings with different difficulty distributions for different stages, the potential of the model is gradually released and the training efficiency is improved. Extensive evaluations show that Compass-Thinker-7B possesses exceptional reasoning potential, and achieves superior performance on mathematics compared to the same-sized RL model. Especially in the challenging AIME2024 evaluation, Compass-Thinker-7B achieves 40% accuracy.

cs.AI↗

arXiv2025

dots.llm1 Technical Report

Mixture of Experts (MoE) models have emerged as a promising paradigm for scaling language models efficiently by activating only a subset of parameters for each input token. In this report, we present dots.llm1, a large-scale MoE model that activates 14B parameters out of a total of 142B parameters, delivering performance on par with state-of-the-art models while reducing training and inference costs. Leveraging our meticulously crafted and efficient data processing pipeline, dots.llm1 achieves performance comparable to Qwen2.5-72B after pretraining on 11.2T high-quality tokens and post-training to fully unlock its capabilities. Notably, no synthetic data is used during pretraining. To foster further research, we open-source intermediate training checkpoints at every one trillion tokens, providing valuable insights into the learning dynamics of large language models.

cs.CL↗

arXiv2025

Establishing Reliability Metrics for Reward Models in Large Language Models

The reward model (RM) that represents human preferences plays a crucial role in optimizing the outputs of large language models (LLMs), e.g., through reinforcement learning from human feedback (RLHF) or rejection sampling. However, a long challenge for RM is its uncertain reliability, i.e., LLM outputs with higher rewards may not align with actual human preferences. Currently, there is a lack of a convincing metric to quantify the reliability of RMs. To bridge this gap, we propose the \textit{\underline{R}eliable at \underline{$η$}} (RETA) metric, which directly measures the reliability of an RM by evaluating the average quality (scored by an oracle) of the top $η$ quantile responses assessed by an RM. On top of RETA, we present an integrated benchmarking pipeline that allows anyone to evaluate their own RM without incurring additional Oracle labeling costs. Extensive experimental studies demonstrate the superior stability of RETA metric, providing solid evaluations of the reliability of various publicly available and proprietary RMs. When dealing with an unreliable RM, we can use the RETA metric to identify the optimal quantile from which to select the responses.

cs.AI↗

arXiv2024

Large deviations for generalized backward stochastic differential equations

This work concerns generalized backward stochastic differential equations, which are coupled with a family of reflecting diffusion processes. First of all, we establish the large deviation principle for forward stochastic differential equations with reflecting boundaries under weak monotonicity conditions. Then based on the obtained result and the contraction principle, the large deviation principle for the generalized backward stochastic differential equations is proved. As a by-product, we obtain a limit result about parabolic partial differential equations with the nonlinear Neumann boundary conditions.

math.PR↗

arXiv2021

Ultrafast Measurements of the Interfacial Spin Seebeck Effect in Au and Rare-Earth Iron Garnet Bilayers

We investigate picosecond spin-currents across Au/iron-garnet interfaces in response to ultrafast laser heating of the electrons in the Au film. In the picoseconds after optical heating, interfacial spin currents occur due to an interfacial temperature difference between electrons in the metal and magnons in the insulator. We report measurements of this interfacial longitudinal spin Seebeck effect between Au and rare-earth iron-garnet insulators, i.e. RE$_3$ Fe$_5$O$_{12}$, where RE is Y, Eu, Tb, Tm. We use time domain thermoreflectance (TDTR) measurements to characterize the thermal response of the bilayer to ultrafast optical heating. We use time-resolved magneto-optic Kerr effect (TR-MOKE) measurements of the Au layer to measure the time-evolution of spin accumulation in the Au film. We observe a spin Seebeck effect between Au/TmIG that is three times larger than for an Au/YIG bilayer. The interfacial thermal conductance between electrons in the Au and magnons in the TmIG layer is ~ 3 $\frac{MW}{m^2 K}$.

cond-mat.mtrl-sci↗

arXiv2021

Imitate TheWorld: A Search Engine Simulation Platform

Recent E-commerce applications benefit from the growth of deep learning techniques. However, we notice that many works attempt to maximize business objectives by closely matching offline labels which follow the supervised learning paradigm. This results in models obtain high offline performance in terms of Area Under Curve (AUC) and Normalized Discounted Cumulative Gain (NDCG), but cannot consistently increase the revenue metrics such as purchases amount of users. Towards the issues, we build a simulated search engine AESim that can properly give feedback by a well-trained discriminator for generated pages, as a dynamic dataset. Different from previous simulation platforms which lose connection with the real world, ours depends on the real data in AliExpress Search: we use adversarial learning to generate virtual users and use Generative Adversarial Imitation Learning (GAIL) to capture behavior patterns of users. Our experiments also show AESim can better reflect the online performance of ranking models than classic ranking metrics, implying AESim can play a surrogate of AliExpress Search and evaluate models without going online.

cs.AI↗

arXiv2021

On-Demand and Lightweight Knowledge Graph Generation -- a Demonstration with DBpedia

Modern large-scale knowledge graphs, such as DBpedia, are datasets which require large computational resources to serve and process. Moreover, they often have longer release cycles, which leads to outdated information in those graphs. In this paper, we present DBpedia on Demand -- a system which serves DBpedia resources on demand without the need to materialize and store the entire graph, and which even provides limited querying functionality.

cs.IR↗

arXiv2020

AliExpress Learning-To-Rank: Maximizing Online Model Performance without Going Online

Learning-to-rank (LTR) has become a key technology in E-commerce applications. Most existing LTR approaches follow a supervised learning paradigm from offline labeled data collected from the online system. However, it has been noticed that previous LTR models can have a good validation performance over offline validation data but have a poor online performance, and vice versa, which implies a possible large inconsistency between the offline and online evaluation. We investigate and confirm in this paper that such inconsistency exists and can have a significant impact on AliExpress Search. Reasons for the inconsistency include the ignorance of item context during the learning, and the offline data set is insufficient for learning the context. Therefore, this paper proposes an evaluator-generator framework for LTR with item context. The framework consists of an evaluator that generalizes to evaluate recommendations involving the context, and a generator that maximizes the evaluator score by reinforcement learning, and a discriminator that ensures the generalization of the evaluator. Extensive experiments in simulation environments and AliExpress Search online system show that, firstly, the classic data-based metrics on the offline dataset can show significant inconsistency with online performance, and can even be misleading. Secondly, the proposed evaluator score is significantly more consistent with the online performance than common ranking metrics. Finally, as the consequence, our method achieves a significant improvement (\textgreater$2\%$) in terms of Conversion Rate (CR) over the industrial-level fine-tuned model in online A/B tests.

cs.LG↗

arXiv2019

Topological Hall effect at above room temperature in heterostructures composed of a magnetic insulator and a heavy metal

Non-volatile memory and computing technology rely on efficient read and write of ultra-tiny information carriers that do not wear out. Magnetic skyrmions are emerging as a potential carrier since they are topologically robust nanoscale spin textures that can be manipulated with ultralow current density. To date, most of skyrmions are reported in metallic films, which suffer from additional Ohmic loss and thus high energy dissipation. Therefore, skyrmions in magnetic insulators are of technological importance for low-power information processing applications due to their low damping and the absence of Ohmic loss. Moreover, they attract fundamental interest in studying various magnon-skyrmion interactions11. Skyrmions have been observed in one insulating material Cu2OSeO3 at cryogenic temperatures, where they are stabilized by bulk Dzyaloshinskii-Moriya interaction. Here, we report the observation of magnetic skyrmions that survive above room temperature in magnetic insulator/heavy metal heterostructures, i.e., thulium iron garnet/platinum. The presence of these skyrmions results from the Dzyaloshinskii-Moriya interaction at the interface and is identified by the emergent topological Hall effect. Through tuning the magnetic anisotropy via varying temperature, we observe skyrmions in a large window of external magnetic field and enhanced stability of skyrmions in the easy-plane anisotropy regime. Our results will help create a new platform for insulating skyrmion-based room temperature low dissipation spintronic applications.

cond-mat.mtrl-sci↗

arXiv2019

Exploring interfacial exchange coupling and sublattice effect in heavy metal/ferrimagnetic insulator heterostructures using Hall measurements, x-ray magnetic circular dichroism, and neutron reflectometry

We use temperature-dependent Hall measurements to identify contributions of spin Hall, magnetic proximity, and sublattice effects to the anomalous Hall signal in heavy metal/ferrimagnetic insulator heterostructures with perpendicular magnetic anisotropy. This approach enables detection of both the magnetic proximity effect onset temperature and the magnetization compensation temperature and provides essential information regarding the interfacial exchange coupling. Onset of a magnetic proximity effect yields a local extremum in the temperature-dependent anomalous Hall signal, which occurs at higher temperature as magnetic insulator thickness increases. This magnetic proximity effect onset occurs at much higher temperature in Pt than W. The magnetization compensation point is identified by a sharp anomalous Hall sign change and divergent coercive field. We directly probe the magnetic proximity effect using x-ray magnetic circular dichroism and polarized neutron reflectometry, which reveal an antiferromagnetic coupling between W and the magnetic insulator. Finally, we summarize the exchange-coupling configurations and the anomalous Hall-effect sign of the magnetized heavy metal in various heavy metal/magnetic insulator heterostructures.

cond-mat.mtrl-sci↗

arXiv2018

Dirac-Surface-State Modulated Spin Dynamics in a Ferrimagnetic Insulator at Room Temperature

This work demonstrates dramatically modified spin dynamics of magnetic insulator (MI) by the spin-momentum locked Dirac surface states of the adjacent topological insulator (TI) which can be harnessed for spintronic applications. As the Bi-concentration x is systematically tuned in 5 nm thick (BixSb1-x)2Te3 TI film, the weight of the surface relative to bulk states peaks at x = 0.32 when the chemical potential approaches the Dirac point. At this concentration, the Gilbert damping constant of the precessing magnetization in 10 nm thick Y3Fe5O12 MI film in the MI/TI heterostructures is enhanced by an order of magnitude, the largest among all concentrations. In addition, the MI acquires additional strong magnetic anisotropy that favors the in-plane orientation with similar Bi-concentration dependence. These extraordinary effects of the Dirac surface states distinguish TI from other materials such as heavy metals in modulating spin dynamics of the neighboring magnetic layer.

cond-mat.mtrl-sci↗

arXiv2018

Effect of distance on photoluminescence quenching and proximity-induced spin-orbit coupling in graphene-WSe2 heterostructures

Spin-orbit coupling (SOC) in graphene can be greatly enhanced by proximity coupling it to transition metal dichalcogenides (TMDs) such as WSe2. We find that the strength of the acquired SOC in graphene depends on the stacking order of the heterostructures when using hexagonal boron nitride (h-BN) as the capping layer, i.e., SiO2/graphene/WSe2/h-BN exhibiting stronger SOC than SiO2/WSe2/graphene/h-BN. We utilize photoluminescence (PL) as an indicator to characterize the interaction between graphene and monolayer WSe2 grown by chemical vapor deposition. We observe much stronger PL quenching in the SiO2/graphene/WSe2/h-BN stack than in the SiO2/WSe2/graphene/h-BN stack, and correspondingly a much larger weak antilocalization (WAL) effect or stronger induced SOC in the former than in the latter. We attribute these two effects to the interlayer distance between graphene and WSe2, which depends on whether graphene is in immediate contact with h-BN. Our observations and hypothesis are further supported by first-principles calculations which reveal a clear difference in the interlayer distance between graphene and WSe2 in these two stacks.

cond-mat.mtrl-sci↗

arXiv2018

Role of dimensional crossover on spin-orbit torque efficiency in magnetic insulator thin films

Magnetic insulators (MIs) attract tremendous interest for spintronic applications due to low Gilbert damping and absence of Ohmic loss. Magnetic order of MIs can be manipulated and even switched by spin-orbit torques (SOTs) generated through spin Hall effect and Rashba-Edelstein effect in heavy metal/MI bilayers. SOTs on MIs are more intriguing than magnetic metals since SOTs cannot be transferred to MIs through direct injection of electron spins. Understanding of SOTs on MIs remains elusive, especially how SOTs scale with the film thickness. Here, we observe the critical role of dimensionality on the SOT efficiency by systematically studying the MI layer thickness dependent SOT efficiency in tungsten/thulium iron garnet (W/TmIG) bilayers. We first show that the TmIG thin film evolves from two-dimensional to three-dimensional magnetic phase transitions as the thickness increases, due to the suppression of long-wavelength thermal fluctuation. Then, we report the significant enhancement of the measured SOT efficiency as the thickness increases. We attribute this effect to the increase of the magnetic moment density in concert with the suppression of thermal fluctuations. At last, we demonstrate the current-induced SOT switching in the W/TmIG bilayers with a TmIG thickness up to 15 nm. The switching current density is comparable with those of heavy metal/ferromagnetic metal cases. Our findings shed light on the understanding of SOTs in MIs, which is important for the future development of ultrathin MI-based low-power spintronics.

cond-mat.mes-hall↗

arXiv2017

Strong electron-hole symmetric Rashba spin-orbit coupling in graphene/monolayer transition metal dichalcogenide heterostructures

Despite its extremely weak intrinsic spin-orbit coupling (SOC), graphene has been shown to acquire considerable SOC by proximity coupling with exfoliated transition metal dichalcogenides (TMDs). Here we demonstrate strong induced Rashba SOC in graphene that is proximity coupled to a monolayer TMD film, MoS2 or WSe2, grown by chemical vapor deposition with drastically different Fermi level positions. Graphene/TMD heterostructures are fabricated with a pickup-transfer technique utilizing hexagonal boron nitride, which serves as a flat template to promote intimate contact and therefore a strong interfacial interaction between TMD and graphene as evidenced by quenching of the TMD photoluminescence. We observe strong induced graphene SOC that manifests itself in a pronounced weak anti-localization (WAL) effect in the graphene magnetoconductance. The spin relaxation rate extracted from the WAL analysis varies linearly with the momentum scattering time and is independent of the carrier type. This indicates a dominantly Dyakonov-Perel spin relaxation mechanism caused by the induced Rashba SOC. Our analysis yields a Rashba SOC energy of ~1.5 meV in graphene/WSe2 and ~0.9 meV in graphene/MoS2, respectively. The nearly electron-hole symmetric nature of the induced Rashba SOC provides a clue to possible underlying SOC mechanisms.

cond-mat.mtrl-sci↗

arXiv2017

Above 400 K Robust Perpendicular Ferromagnetic Phase in a Topological Insulator

The quantum anomalous Hall effect (QAHE) that emerges under broken time-reversal symmetry in topological insulators (TI) exhibits many fascinating physical properties for potential applications in nano-electronics and spintronics. However, in transition-metal doped TI, the only experimentally demonstrated QAHE system to date, the effect is lost at practically relevant temperatures. This constraint is imposed by the relatively low Curie temperature (Tc) and inherent spin disorder associated with the random magnetic dopants. Here we demonstrate drastically enhanced Tc by exchange coupling TI to Tm3Fe5O12, a high-Tc magnetic insulator with perpendicular magnetic anisotropy. Signatures that the TI surface states acquire robust ferromagnetism are revealed by distinct squared anomalous Hall hysteresis loops at 400 K. Point-contact Andreev reflection spectroscopy confirms that the TI surface is indeed spin-polarized. The greatly enhanced Tc, absence of spin disorder, and perpendicular anisotropy are all essential to the occurrence of the QAHE at high temperatures.

cond-mat.mtrl-sci↗

arXiv2017

Deficiency of the Bulk Spin Hall Effect Model for Spin-Orbit Torques in Magnetic Insulator/Heavy Metal Heterostructures

Electrical currents in a magnetic insulator/heavy metal heterostructure can induce two simultaneous effects, namely, spin Hall magnetoresistance (SMR) on the heavy metal side and spin-orbit torques (SOTs) on the magnetic insulator side. Within the framework of the pure spin current model based on the bulk spin Hall effect (SHE), the ratio of the spin Hall-induced anomalous Hall effect (SH-AHE) to SMR should be equal to the ratio of the field-like torque (FLT) to damping-like torque (DLT). We perform a quantitative study of SMR, SH-AHE, and SOTs in a series of thulium iron garnet/platinum or Tm3Fe5O12/Pt heterostructures with different Tm3Fe5O12 thicknesses, where Tm3Fe5O12 is a ferrimagnetic insulator with perpendicular magnetic anisotropy. We find the ratio between measured effective fields of FLT and DLT is at least 2 times larger than the ratio of the SH-AHE to SMR. In addition, the bulk SHE model grossly underestimates the spin torque efficiency of FLT. Our results reveal deficiencies of the bulk SHE model and also address the importance of interfacial effects such as the Rashba and magnetic proximity effects in magnetic insulator/heavy metal heterostructures.

cond-mat.mtrl-sci↗

arXiv2016

Anomalous Hall hysteresis in Tm3Fe5O12/Pt with strain-induced perpendicular magnetic anisotropy

We demonstrate robust interface strain-induced perpendicular magnetic anisotropy in atomically flat ferrimagnetic insulator Tm3Fe5O12 films grown with pulsed laser deposition on substituted-Gd3Ga5O12 substrate which maximizes the tensile strain at the interface. In bilayers consisting of Pt and TIG, we observe large squared Hall hysteresis loops over a wide range of thicknesses of Pt at room temperature. When a thin Cu layer is inserted between Pt and TIG, the Hall hysteresis magnitude decays but stays finite as the thickness of Cu increases up to 5 nm. However, if the Cu layer is placed atop Pt instead, the Hall hysteresis magnitude is consistently larger than when the Cu layer with the same thickness is inserted in between for all Cu thicknesses. These results suggest that both the proximity-induced ferromagnetism and spin current contribute to the anomalous Hall effect.

cond-mat.mtrl-sci↗