Search

We present a phase-space electronic Hamiltonian $\hat{H}_{PS}$ (parameterized by both nuclear position $\mathbf{X}$ and momentum $\mathbf{P}$) that boosts each electron into the moving frame of the nuclei that are closest in real space -- without presuming the existence of an atomic orbital basis. We show that $(i)$ quantum-classical dynamics along such a Hamiltonian maintains momentum conservation and $(ii)$ diagonalizing such a Hamiltonian can recover the electronic momentum and electronic current density reasonably well. In conjunction with other reports in the literature that such a phase-space approach can also recover vibrational circular dichroism (VCD) spectra, we submit that the present phase-space approach offers a testable and powerful approach to post-Born-Oppenheimer electronic structure theory. Moreover, the approach is inexpensive and can be immediately applied to simulations of chiral induced spin selectivity experiments (where the transfer of angular momentum between nuclei and electrons is considered critical).

Large language models (LLMs) have revolutionized Natural Language Processing (NLP) by minimizing the need for complex feature engineering. However, the application of LLMs in specialized domains like biopharmaceuticals and chemistry remains largely unexplored. These fields are characterized by intricate terminologies, specialized knowledge, and a high demand for precision areas where general purpose LLMs often fall short. In this study, we introduce PharmaGPT, a suite of domain specilized LLMs with 13 billion and 70 billion parameters, specifically trained on a comprehensive corpus tailored to the Bio-Pharmaceutical and Chemical domains. Our evaluation shows that PharmaGPT surpasses existing general models on specific-domain benchmarks such as NAPLEX, demonstrating its exceptional capability in domain-specific tasks. Remarkably, this performance is achieved with a model that has only a fraction, sometimes just one-tenth-of the parameters of general-purpose large models. This advancement establishes a new benchmark for LLMs in the bio-pharmaceutical and chemical fields, addressing the existing gap in specialized language modeling. It also suggests a promising path for enhanced research and development, paving the way for more precise and effective NLP applications in these areas.

In recent years, large language models(LLMs) have attracted significant attention due to their exceptional performance across a multitude of natural language process tasks, and have been widely applied in various fields. However, the application of large language models in the Intellectual Property (IP) domain is challenging due to the strong need for specialized knowledge, privacy protection, processing of extremely long text in this field. In this technical report, we present for the first time a low-cost, standardized procedure for training IP-oriented LLMs, meeting the unique requirements of the IP domain. Using this standard process, we have trained the PatentGPT series models based on open-source pretrained models. By evaluating them on the open-source IP-oriented benchmark MOZIP, our domain-specific LLMs outperforms GPT-4, indicating the effectiveness of the proposed training procedure and the expertise of the PatentGPT models in the IP domain. Remarkably, our model surpassed GPT-4 on the 2019 China Patent Agent Qualification Examination, scoring 65 and matching human expert levels. Additionally, the PatentGPT model, which utilizes the SMoE architecture, achieves performance comparable to that of GPT-4 in the IP domain and demonstrates a better cost-performance ratio on long-text tasks, potentially serving as an alternative to GPT-4 within the IP domain., Comment: 19 pages, 9 figures

Robotic branch pruning is a significantly growing research area to cope with the shortage of labor force in the context of agriculture. One fundamental requirement in robotic pruning is the perception of detailed geometry and topology of branches. However, the point clouds obtained in agricultural settings often exhibit incompleteness due to several constraints, thereby restricting the accuracy of downstream robotic pruning. In this work, we addressed the issue of point cloud quality through a simulation-based deep neural network, leveraging a Real-to-Simulation (Real2Sim) data generation pipeline that not only eliminates the need for manual parameterization but also guarantees the realism of simulated data. The simulation-based neural network was applied to jointly perform point cloud completion and skeletonization on real-world partial branches, without additional real-world training. The Sim2Real qualitative completion and skeletonization results showed the model's remarkable capability for geometry reconstruction and topology prediction. Additionally, we quantitatively evaluated the Sim2Real performance by comparing branch-level trait characterization errors using raw incomplete data and complete data. The Mean Absolute Error (MAE) reduced by 75% and 8% for branch diameter and branch angle estimation, respectively, using the best complete data, which indicates the effectiveness of the Real2Sim data in a zero-shot generalization setting. The characterization improvements contributed to the precision and efficacy of robotic branch pruning., Comment: Submitted to IROS2024

Small-scale robots hold great potential for targeted cargo delivery in minimally-inv asive medicine. However, current robots often face challenges to locomote efficiently on slip pery biological tissue surfaces, especially when loaded with heavy cargos. Here, we report a magnetic millirobot that can walk on rough and slippery biological tissues by anchoring itself on the soft tissue surface alternatingly with two feet and reciprocally rotating the body to mov e forward. We experimentally studied the locomotion, validated it with numerical simulations and optimized the actuation parameters to fit various terrains and loading conditions. Further more, we developed a permanent magnet set-up to enable wireless actuation within a huma n-scale volume which allows precise control of the millirobot to follow complex trajectories, cl imb vertical walls, and carry cargo up to four times of its own weight. Upon reaching the targ et location, it performs a deployment sequence to release the liquid drug into tissues. The ro bust gait of our millirobot on rough biological terrains, combined with its heavy load capacity, make it a versatile and effective miniaturized vehicle for targeted cargo delivery., Comment: 15 pages

On-demand controlled drug delivery is essential for the treatment of a wide range of chronic diseases. As the drug is released at the time when required, its efficacy is boosted and the side effects are minimized. However, so far, drug delivery devices often rely on the passive diffusion process for a sustained release, which is slow and uncontrollable. Here, we present a miniaturized microfluidic device for wirelessly controlled ultrafast active drug delivery, driven by an oscillating solid-liquid interface. The oscillation generates acoustic streaming in the drug reservoir, which opens an elastic valve to deliver the drug. High-speed microscopy reveals the fast response of the valve on the order of 1 ms, which is more than three orders of magnitude faster than the start-of-the-art. The amount of the released drug exhibits a linear relationship with the working time and the electric power applied to the ultrasonic resonator. The trigger of the release is wirelessly controlled via a magnetic field, and the system shows stable output in a continuous experiment for two weeks. The integrated system shows great promise as a long-term controlled drug delivery implant for chronic diseases., Comment: 19 pages, 6 figures, 1 table

We present 4,987 white dwarf (WD) candidates selected from matched stars between the multi-band imaging datasets of Subaru Hyper Suprime-Cam (HSC) Survey and SDSS in the Stripe82 region covering about 165 deg$^2$. We first select WD candidates from the "reduced proper motion" diagram that is obtained by combining the apparent magnitude in the range $i=19$ -- 24 and the proper motion measured by comparing the astrometric positions of each object between the two datasets over about 14 yr time baseline. We refine the WD candidates by fitting blackbody and template WD atmosphere models to HSC photometries for each candidate, enabling the estimation of photometric distance and tangential velocity ($v_{\rm t}$) with respect to the Sun. The deep HSC data allows us to identify low-temperature ($<4000$ K) and faint WD candidates down to absolute magnitude, $M_{\rm bol}\simeq 17$. We evaluate the selection function of our WD candidates using the mock catalogue for spatial and kinematic distributions of WDs in the (thin and thick) disc and halo regions based on the standard Milky Way model. We construct the samples of disc and halo WD candidates by selecting WDs with the cuts of tangential velocity, $40

Modern electronic structure theory is built around the Born-Oppenheimer approximation and the construction of an electronic Hamiltonian H_{el}(X) that depends on the nuclear position X (and not the nuclear momentum P). In this article, using the well-known theory of electron translation (Gamma') and rotational (Gamma'') factors to couple electronic transitions to nuclear motion, we construct a practical phase-space electronic Hamiltonian that depends on both nuclear position and momentum, H_{PS}(X,P). While classical Born-Oppenheimer dynamics that run along the eigensurfaces of the operator H_{el}(X) can recover many nuclear properties correctly, we present some evidence that motion along the eigensurfaces of H_{PS}(X,P) can better capture both nuclear and electronic properties (including the elusive electronic momentum studied by Nafie). Moreover, only the latter (as opposed to the former) conserves the total linear and angular momentum in general.

Within the context of FSSH dynamics, one often wishes to remove the angular component of the derivative coupling between states $\left|J\right>$ and $\left|K\right>$. In a set of previous papers, Truhlar {\em et al.} posited one approach for such a removal based on direct projection, while we isolated a second approach by constructing and differentiating rotationally invariant basis. Unfortunately, neither approach was able to demonstrate a {\em one-electron operator} $\hat{O}$ whose matrix element $\left$ was the angular component of the derivative coupling. Here, we show that a one-electron operator can in fact be constructed efficiently in a semi-local fashion. The present results yield physical insight into designing new surface hopping algorithms and be of immediate use for FSSH calculations.

We analyze the photometric data in the Wide layer of the Hyper Suprime-Cam Subaru Strategic Program (HSC-SSP) over $\sim 1,200$ deg$^{2}$ to uncover new halo substructures beyond the distance, $D_{\odot}\sim$ 30 kpc, from the Sun. For this purpose, we develop an isochrone filter for an old, metal-poor stellar system to extract the faint main-sequence stars at a range of distances. With this method, we detect, not only the previously discovered substructures such as the Orphan Stream, but also the new overdensity toward Bo\"otes at about $D_{\odot}\sim$ 60 kpc and the new stream-like feature toward Pisces at around $D_{\odot}\sim$ 60 kpc. It has been suggested that a small-scale overdensity exists in this direction of Pisces (the so-called Pisces Overdensity), but our results show that the overdensity is widely spread with a tidally elongated feature. Combining our results with the ongoing Hyper Suprime-Cam narrow-band survey and the near-future spectroscopic survey with Prime Focus Spectrograph (PFS) will allow us to place strong constraints on the origin of these halo substructures., Comment: 16 pages, 11 figures, 1 table, accepted for publication in PASJ

Cryptocurrency exchange platforms such as Coinbase, Binance, enable users to purchase and sell cryptocurrencies conveniently just like trading stocks/commodities. However, because of the nature of blockchain, when a user withdraws coins (i.e., transfers coins to an external on-chain account), all future transactions can be learned by the platform. This is in sharp contrast to conventional stock exchange where all external activities of users are always hidden from the platform. Since the platform knows highly sensitive user private information such as passport number, bank information etc, linking all (on-chain) transactions raises a serious privacy concern about the potential disastrous data breach in those cryptocurrency exchange platforms. In this paper, we propose a cryptocurrency exchange that restores user anonymity for the first time. To our surprise, the seemingly well-studied privacy/anonymity problem has several new challenges in this setting. Since the public blockchain and internal transaction activities naturally provide many non-trivial leakages to the platform, internal privacy is not only useful in the usual sense but also becomes necessary for regaining the basic anonymity of user transactions. We also ensure that the user cannot double spend, and the user has to properly report accumulated profit for tax purposes, even in the private setting. We give a careful modeling and efficient construction of the system that achieves constant computation and communication overhead (with only simple cryptographic tools and rigorous security analysis); we also implement our system and evaluate its practical performance., Comment: 27 pages, 8 figures, 2 tables. To be published in NDSS'24. This is the full version of the conference paper

For a system without spin-orbit coupling, the (i) nuclear plus electronic linear momentum and (ii) nuclear plus orbital electronic angular momentum are good quantum numbers. Thus, when a molecular system undergoes a nonadiabatic transition, there should be no change in the total linear or angular momentum. Now, the standard surface hopping algorithm ignores the electronic momentum and indirectly equates the momentum of the nuclear degrees of freedom to the total momentum. However, even with this simplification, the algorithm still does not conserve either the nuclear linear or the nuclear angular momenta. Here, we show that one way to address these failures is to dress the derivative couplings (i.e. the hopping directions) in two ways: (i) we disallow changes in the nuclear linear momentum by working in a translating basis (which is well known and leads to electron translation factors [ETFs]); and (ii) we disallow changes in the nuclear angular momentum by working in a basis that rotates around the center of mass (which is not well-known and leads to a novel, rotationally removable component of the derivative coupling that we will call electron rotation factors [ERFs] below, cf. Eq. 96). The present findings should be helpful in the short term as far as interpreting surface hopping calculations for singlet systems (without spin) and then developing new surface hopping algorithm in the long term for systems where one cannot ignore the electronic orbital and/or spin angular momentum.

A surge of interest has emerged in utilizing Transformers in diverse vision tasks owing to its formidable performance. However, existing approaches primarily focus on optimizing internal model architecture designs that often entail significant trial and error with high burdens. In this work, we propose a new paradigm dubbed Decision Stream Calibration that boosts the performance of general Vision Transformers. To achieve this, we shed light on the information propagation mechanism in the learning procedure by exploring the correlation between different tokens and the relevance coefficient of multiple dimensions. Upon further analysis, it was discovered that 1) the final decision is associated with tokens of foreground targets, while token features of foreground target will be transmitted into the next layer as much as possible, and the useless token features of background area will be eliminated gradually in the forward propagation. 2) Each category is solely associated with specific sparse dimensions in the tokens. Based on the discoveries mentioned above, we designed a two-stage calibration scheme, namely ViT-Calibrator, including token propagation calibration stage and dimension propagation calibration stage. Extensive experiments on commonly used datasets show that the proposed approach can achieve promising results. The source codes are given in the supplements., Comment: At present, the paper involves internal projects of the company, and it is not convenient to publish it temporarily, so the article needs to be withdrawn temporarily

Consistency between the exchange-correlation (xc) functional used during pseudopotential construction and planewave-based electronic structure calculations is important for an accurate and reliable description of the structure and properties of condensed-phase systems. In this work, we present a general scheme for constructing pseudopotentials with range-separated hybrid (RSH) xc functionals based on the solution of the all-electron radial integro-differential equation for a spherical atomic configuration. As proof-of-principle, we demonstrate pseudopotential construction with the PBE, PBE0, HSE, and sRSH (based on LC-$\omega$PBE0) xc functionals for a select set of atoms, and then investigate the importance of pseudopotential consistency when computing band gaps, equilibrium lattice parameters, bulk moduli, and atomization energies of several solid-state systems. In doing so, we find that pseudopotential consistency errors (PSCE) tend to be systematic and can be as large as $1.4\%$ when computing these properties.

Recent proposed DETR variants have made tremendous progress in various scenarios due to their streamlined processes and remarkable performance. However, the learned queries usually explore the global context to generate the final set prediction, resulting in redundant burdens and unfaithful results. More specifically, a query is commonly responsible for objects of different scales and positions, which is a challenge for the query itself, and will cause spatial resource competition among queries. To alleviate this issue, we propose Team DETR, which leverages query collaboration and position constraints to embrace objects of interest more precisely. We also dynamically cater to each query member's prediction preference, offering the query better scale and spatial priors. In addition, the proposed Team DETR is flexible enough to be adapted to other existing DETR variants without increasing parameters and calculations. Extensive experiments on the COCO dataset show that Team DETR achieves remarkable gains, especially for small and large objects. Code is available at \url{https://github.com/horrible-dong/TeamDETR}.

We use the Subaru Hyper Suprime-Cam (HSC) data to study structural parameters and systemic proper motion of the Sextans dwarf spheroidal galaxy at the heliocentric distance of 86 kpc, which is one of the most important targets for studies of dark matter nature and galaxy formation physics. Thanks to the superb image quality and wide area coverage of the Sextans field, the HSC data enables a secure selection of member star candidates based on the colour-magnitude cut, yielding about 10,000 member candidates at magnitudes down to $i\sim 24$. We use a likelihood analysis of the two-dimensional distribution of stars to estimate the structural parameters of Sextans taking into account the contamination of foreground halo stars in the Milky Way, and find that the member star distribution is well-fitted by an elliptical King profile with ellipticity $\epsilon \simeq 0.25$ and the core and tidal radii of $R_c=(368.4\pm 8.5)$ pc and $R_t=(2.54\pm 0.046)$ kpc, respectively. Then using the two HSC datasets of 2.66 years time baseline on average, we find the systemic proper motions of Sextans to be $(\mu_\alpha, \mu_\delta)=(-0.448\pm{0.075},0.083\pm{0.078})$ $\mathrm{mas}\ \mathrm{yr}^{-1}$, which is consistent with some of the previous works using the $Gaia$ data of relatively bright member stars in Sextans. Thus, our results give a demonstration that a ground-based, large-aperture telescope data which covers a wide solid angle of the sky and has a long time baseline, such as the upcoming LSST data, can be used to study systemic proper motions of dwarf galaxies., Comment: 14 pages, 15 figures

We investigate how the Berry force (i.e. the pseudo-magnetic force operating on nuclei as induced by electronic degeneracy and spin-orbit coupling (SOC)) might modify a post-transition state bifurcation (PTSB) reaction path and affect product selectivity for situations when multiple products share the same transition state. To estimate the magnitude of this effect, Langevin dynamics are performed on a model system with a valley-ridge inflection (VRI) point in the presence of a magnetic field (that mimics the Berry curvature). We also develop an analytic model for such selectivity that depends on key parameters such as the surface topology, the magnitude of the Berry force, and the nuclear friction. Within this dynamical model, static electronic structure calculations (at the level of generalized Hartree-Fock with spin-orbit coupling (GHF+SOC) theory) suggest that electronic-spin induced Berry force effects may indeed lead to noticeable changes in methoxy radical isomerization.

To acquire a snapshot spectral image, coded aperture snapshot spectral imaging (CASSI) is proposed. A core problem of the CASSI system is to recover the reliable and fine underlying 3D spectral cube from the 2D measurement. By alternately solving a data subproblem and a prior subproblem, deep unfolding methods achieve good performance. However, in the data subproblem, the used sensing matrix is ill-suited for the real degradation process due to the device errors caused by phase aberration, distortion; in the prior subproblem, it is important to design a suitable model to jointly exploit both spatial and spectral priors. In this paper, we propose a Residual Degradation Learning Unfolding Framework (RDLUF), which bridges the gap between the sensing matrix and the degradation process. Moreover, a Mix$S^2$ Transformer is designed via mixing priors across spectral and spatial to strengthen the spectral-spatial representation capability. Finally, plugging the Mix$S^2$ Transformer into the RDLUF leads to an end-to-end trainable neural network RDLUF-Mix$S^2$. Experimental results establish the superior performance of the proposed method over existing ones., Comment: CVPR 2023

Deep learning has recently achieved remarkable performance in image classification tasks, which depends heavily on massive annotation. However, the classification mechanism of existing deep learning models seems to contrast to humans' recognition mechanism. With only a glance at an image of the object even unknown type, humans can quickly and precisely find other same category objects from massive images, which benefits from daily recognition of various objects. In this paper, we attempt to build a generalizable framework that emulates the humans' recognition mechanism in the image classification task, hoping to improve the classification performance on unseen categories with the support of annotations of other categories. Specifically, we investigate a new task termed Comparison Knowledge Translation (CKT). Given a set of fully labeled categories, CKT aims to translate the comparison knowledge learned from the labeled categories to a set of novel categories. To this end, we put forward a Comparison Classification Translation Network (CCT-Net), which comprises a comparison classifier and a matching discriminator. The comparison classifier is devised to classify whether two images belong to the same category or not, while the matching discriminator works together in an adversarial manner to ensure whether classified results match the truth. Exhaustive experiments show that CCT-Net achieves surprising generalization ability on unseen categories and SOTA performance on target categories., Comment: Accepted by IJCAI 2022; Adding Supplementary Materials

We show that the Berry force as computed by an approximate, mean-field electronic structure can be meaningful if properly interpreted. In particular, for a model Hamiltonian representing a molecular system with an even number of electrons interacting via a two-body (Hubbard) interaction and a spin-orbit coupling, we show that a meaningful nonzero Berry force emerges whenever there is spin unrestriction--even though the Hamiltonian is real-valued and formally the on-diagonal single-surface Berry force must be zero. Moreover, if properly applied, this mean-field Berry force yields roughly the correct asymptotic motion for scattering through an avoided crossing. That being said, within the context of a ground-state calculation, several nuances do arise as far interpreting the Berry force correctly, and as a practical matter, the Berry force diverges near the Coulson-Fisher point (which can lead to numerical instabilities). We do not address magnetic fields here.

Quantification of behavior is critical in applications ranging from neuroscience, veterinary medicine and animal conservation efforts. A common key step for behavioral analysis is first extracting relevant keypoints on animals, known as pose estimation. However, reliable inference of poses currently requires domain knowledge and manual labeling effort to build supervised models. We present a series of technical innovations that enable a new method, collectively called SuperAnimal, to develop unified foundation models that can be used on over 45 species, without additional human labels. Concretely, we introduce a method to unify the keypoint space across differently labeled datasets (via our generalized data converter) and for training these diverse datasets in a manner such that they don't catastrophically forget keypoints given the unbalanced inputs (via our keypoint gradient masking and memory replay approaches). These models show excellent performance across six pose benchmarks. Then, to ensure maximal usability for end-users, we demonstrate how to fine-tune the models on differently labeled data and provide tooling for unsupervised video adaptation to boost performance and decrease jitter across frames. If the models are fine-tuned, we show SuperAnimal models are 10-100$\times$ more data efficient than prior transfer-learning-based approaches. We illustrate the utility of our models in behavioral classification in mice and gait analysis in horses. Collectively, this presents a data-efficient solution for animal pose estimation., Comment: Models and demos available at http://modelzoo.deeplabcut.org

Quantum Brownian motion, described by the Caldeira-Leggett model, brings insights to understand phenomena and essence of quantum thermodynamics, especially the quantum work and heat associated with their classical counterparts. By employing the phase-space formulation approach, we study the heat distribution of a relaxation process in the quantum Brownian motion model. The analytical result of the characteristic function of heat is obtained at any relaxation time with an arbitrary friction coefficient. By taking the classical limit, such a result approaches the heat distribution of the classical Brownian motion described by the Langevin equation, indicating the quantum-classical correspondence principle for heat distribution. We also demonstrate that the fluctuating heat at any relaxation time satisfies the exchange fluctuation theorem of heat, and its long-time limit reflects complete thermalization of the system. Our research brings justification for the definition of the quantum fluctuating heat via two-point measurements., Comment: 32 pages, 2 figures

The deep learning approach to detecting malicious software (malware) is promising but has yet to tackle the problem of dataset shift, namely that the joint distribution of examples and their labels associated with the test set is different from that of the training set. This problem causes the degradation of deep learning models without users' notice. In order to alleviate the problem, one approach is to let a classifier not only predict the label on a given example but also present its uncertainty (or confidence) on the predicted label, whereby a defender can decide whether to use the predicted label or not. While intuitive and clearly important, the capabilities and limitations of this approach have not been well understood. In this paper, we conduct an empirical study to evaluate the quality of predictive uncertainties of malware detectors. Specifically, we re-design and build 24 Android malware detectors (by transforming four off-the-shelf detectors with six calibration methods) and quantify their uncertainties with nine metrics, including three metrics dealing with data imbalance. Our main findings are: (i) predictive uncertainty indeed helps achieve reliable malware detection in the presence of dataset shift, but cannot cope with adversarial evasion attacks; (ii) approximate Bayesian methods are promising to calibrate and generalize malware detectors to deal with dataset shift, but cannot cope with adversarial evasion attacks; (iii) adversarial evasion attacks can render calibration methods useless, and it is an open problem to quantify the uncertainty associated with the predicted labels of adversarial examples (i.e., it is not effective to use predictive uncertainty to detect adversarial examples)., Comment: Accepted by ACSAC'2021