Search

Optimal input settings vary across users due to differences in motor abilities and personal preferences, which are typically addressed by manual tuning or calibration. Although human-in-the-loop optimization has the potential to identify optimal settings during use, it is rarely applied due to its long optimization process. A more efficient approach would continually leverage data from previous users to accelerate optimization, exploiting shared traits while adapting to individual characteristics. We introduce the concept of Continual Human-in-the-Loop Optimization and a Bayesian optimization-based method that leverages a Bayesian-neural-network surrogate model to capture population-level characteristics while adapting to new users. We propose a generative replay strategy to mitigate catastrophic forgetting. We demonstrate our method by optimizing virtual reality keyboard parameters for text entry using direct touch, showing reduced adaptation times with a growing user base. Our method opens the door for next-generation personalized input systems that improve with accumulated experience.

Supernumerary robotic limbs (SRLs) are robotic structures integrated closely with the user's body, which augment human physical capabilities and necessitate seamless, naturalistic human-machine interaction. For effective assistance in physical tasks, enabling SRLs to hand over objects to humans is crucial. Yet, designing heuristic-based policies for robots is time-consuming, difficult to generalize across tasks, and results in less human-like motion. When trained with proper datasets, generative models are powerful alternatives for creating naturalistic handover motions. We introduce 3HANDS, a novel dataset of object handover interactions between a participant performing a daily activity and another participant enacting a hip-mounted SRL in a naturalistic manner. 3HANDS captures the unique characteristics of SRL interactions: operating in intimate personal space with asymmetric object origins, implicit motion synchronization, and the user's engagement in a primary task during the handover. To demonstrate the effectiveness of our dataset, we present three models: one that generates naturalistic handover trajectories, another that determines the appropriate handover endpoints, and a third that predicts the moment to initiate a handover. In a user study (N=10), we compare the handover interaction performed with our method compared to a baseline. The findings show that our method was perceived as significantly more natural, less physically demanding, and more comfortable., Comment: CHI '25

In order to deeply understand the capability of pretrained language models in text generation and conduct a diagnostic evaluation, we propose TGEA, an error-annotated dataset with multiple benchmark tasks for text generation from pretrained language models (PLMs). We use carefully selected prompt words to guide GPT-2 to generate candidate sentences, from which we select 47K for error annotation. Crowdsourced workers manually check each of these sentences and detect 12k erroneous sentences. We create an error taxonomy to cover 24 types of errors occurring in these erroneous sentences according to the nature of errors with respect to linguistics and knowledge (eg, common sense). For each erroneous span in PLM-generated sentences, we also detect another span that is closely associated with it. Each error is hence manually labeled with comprehensive annotations, including the span of the error, the associated span, minimal correction to the error, the type of the error, and rationale behind the error. Apart from the fully annotated dataset, we also present a detailed description of the data collection procedure, statistics and analysis of the dataset. This is the first dataset with comprehensive annotations for PLM-generated texts, which facilitates the diagnostic evaluation of PLM-based text generation. Furthermore, we use TGEA as a benchmark dataset and propose a series of automatic diagnosis tasks, including error detection, error type classification, associated span detection, error rationale generation, to further promote future study on the automatic error detection and correction on texts generated by pretrained language models., Comment: ACL 2021

Diffusion models are the standard toolkit for generative modelling of 3D atomic systems. However, for different types of atomic systems - such as molecules and materials - the generative processes are usually highly specific to the target system despite the underlying physics being the same. We introduce the All-atom Diffusion Transformer (ADiT), a unified latent diffusion framework for jointly generating both periodic materials and non-periodic molecular systems using the same model: (1) An autoencoder maps a unified, all-atom representations of molecules and materials to a shared latent embedding space; and (2) A diffusion model is trained to generate new latent embeddings that the autoencoder can decode to sample new molecules or materials. Experiments on QM9 and MP20 datasets demonstrate that jointly trained ADiT generates realistic and valid molecules as well as materials, exceeding state-of-the-art results from molecule and crystal-specific models. ADiT uses standard Transformers for both the autoencoder and diffusion model, resulting in significant speedups during training and inference compared to equivariant diffusion models. Scaling ADiT up to half a billion parameters predictably improves performance, representing a step towards broadly generalizable foundation models for generative chemistry. Open source code: https://github.com/facebookresearch/all-atom-diffusion-transformer

In this paper, we systematically investigate the general spin-one dark matter-nucleus interactions within the framework of effective field theories (EFT). We consider both the nonrelativistic (NR) and the relativistic EFT descriptions of the DM interactions with nucleons and quarks. In the NREFT framework, we present a complete list of NR operators for spin-one DM coupling to nucleons and compute their contributions to the DM response functions. Next, we consider all possible leading-order relativistic EFT operators between DM and light quarks and the photon, and perform NR reductions to match them onto the NREFT. We then derive the nuclear scattering rate from these interactions, and employ recent DM direct detection data (from both the nuclear recoil and the Migdal effect) to constrain all these EFT operators and DM electromagnetic properties. We find the elastic nuclear recoil data (from PandaX-4T, XENONnT, LZ, and DarkSide-50) set stringent bounds on the EFT coefficients for a DM mass above a few GeV while the Migdal effect datasets (from PandaX-4T, XENONnT, and DarkSide-50) can probe the DM mass region as small as 20 MeV. Lastly, we construct a UV complete model that can provide a complex spin-one DM candidate, and at the same time generate DM-quark/photon operators discussed in this work., Comment: 35 pages, 9 figures

Affordances, a foundational concept in human-computer interaction and design, have traditionally been explained by direct-perception theories, which assume that individuals perceive action possibilities directly from the environment. However, these theories fall short of explaining how affordances are perceived, learned, refined, or misperceived, and how users choose between multiple affordances in dynamic contexts. This paper introduces a novel affordance theory grounded in Computational Rationality, positing that humans construct internal representations of the world based on bounded sensory inputs. Within these internal models, affordances are inferred through two core mechanisms: feature recognition and hypothetical motion trajectories. Our theory redefines affordance perception as a decision-making process, driven by two components: confidence (the perceived likelihood of successfully executing an action) and predicted utility (the expected value of the outcome). By balancing these factors, individuals make informed decisions about which actions to take. Our theory frames affordances perception as dynamic, continuously learned, and refined through reinforcement and feedback. We validate the theory via thought experiments and demonstrate its applicability across diverse types of affordances (e.g., physical, digital, social). Beyond clarifying and generalizing the understanding of affordances across contexts, our theory serves as a foundation for improving design communication and guiding the development of more adaptive and intuitive systems that evolve with user capabilities., Comment: IUI 2025 Paper

Most studies on baryon number violating (BNV) processes in the literature focus on free or bound nucleons in nuclei, with limited attention given to the decay of bound atoms. Given that hydrogen is the most abundant atom in the universe, it is particularly intriguing to investigate the decay of hydrogen atom as a means to probe BNV interactions. In this study, for the first time, we employ a robust effective field theory (EFT) approach to estimate the decay widths of two-body decays of hydrogen atom into standard model particles, by utilizing the constraints on the EFT cutoff scale derived from conventional nucleon decay processes. We integrate low energy effective field theory (LEFT), chiral perturbation theory (ChPT), and standard model effective field theory (SMEFT) to formulate the decay widths in terms of the LEFT and SMEFT Wilson coefficients (WCs), respectively. By applying the bounds on the WCs from conventional nucleon decays, we provide a conservative estimate on hydrogen BNV decays. Our findings indicate that the bounds on the inverse partial widths of all dominant two-body decays exceed $10^{44}$ years. Among these modes, the least constrained diphoton decay $\Hy\to \gamma\gamma$ might be astrophysically interesting, although the monochromatic photon signal from our Sun is difficult to detect with current near-Earth telescopes., Comment: 13 pages, 3 tables, matched the published version in PLB

In this paper, we present a Neuron Abandoning Attention Flow (NAFlow) method to address the open problem of visually explaining the attention evolution dynamics inside CNNs when making their classification decisions. A novel cascading neuron abandoning back-propagation algorithm is designed to trace neurons in all layers of a CNN that involve in making its prediction to address the problem of significant interference from abandoned neurons. Firstly, a Neuron Abandoning Back-Propagation (NA-BP) module is proposed to generate Back-Propagated Feature Maps (BPFM) by using the inverse function of the intermediate layers of CNN models, on which the neurons not used for decision-making are abandoned. Meanwhile, the cascading NA-BP modules calculate the tensors of importance coefficients which are linearly combined with the tensors of BPFMs to form the NAFlow. Secondly, to be able to visualize attention flow for similarity metric-based CNN models, a new channel contribution weights module is proposed to calculate the importance coefficients via Jacobian Matrix. The effectiveness of the proposed NAFlow is validated on nine widely-used CNN models for various tasks of general image classification, contrastive learning classification, few-shot image classification, and image retrieval.

Over the past few years, vision transformers (ViTs) have consistently demonstrated remarkable performance across various visual recognition tasks. However, attempts to enhance their robustness have yielded limited success, mainly focusing on different training strategies, input patch augmentation, or network structural enhancements. These approaches often involve extensive training and fine-tuning, which are time-consuming and resource-intensive. To tackle these obstacles, we introduce a novel approach named Spatial Autocorrelation Token Analysis (SATA). By harnessing spatial relationships between token features, SATA enhances both the representational capacity and robustness of ViT models. This is achieved through the analysis and grouping of tokens according to their spatial autocorrelation scores prior to their input into the Feed-Forward Network (FFN) block of the self-attention mechanism. Importantly, SATA seamlessly integrates into existing pre-trained ViT baselines without requiring retraining or additional fine-tuning, while concurrently improving efficiency by reducing the computational load of the FFN units. Experimental results show that the baseline ViTs enhanced with SATA not only achieve a new state-of-the-art top-1 accuracy on ImageNet-1K image classification (94.9%) but also establish new state-of-the-art performance across multiple robustness benchmarks, including ImageNet-A (top-1=63.6%), ImageNet-R (top-1=79.2%), and ImageNet-C (mCE=13.6%), all without requiring additional training or fine-tuning of baseline models.

We investigate the impact of effective interactions of dimension-8 (dim-8) operators in the standard model effective field theory (SMEFT) on neutral meson mixing, focusing on the $K^0-\bar K^0$, $B_{d,s}-\bar B_{d,s}$, and $D^0-\bar D^0$ systems. Within the framework of the low energy effective field theory (LEFT), each system is governed by eight dim-6 operators, with four originating at tree level from dim-6 SMEFT operators and the other four from dim-8 SMEFT operators. Notably, in certain UV complete models those dim-8 operators instead of the dim-6 ones are generated at the leading order. Our analysis focuses on those dim-8 operators and includes their one-loop QCD renormalization group running effects. By leveraging the LEFT master formula we impose stringent constraints on the effective scales associated with these dim-8 operators. We find that neutral meson mixing can probe an effective scale up to 50 TeV for some operators, surpassing the constraints imposed on other dim-8 operators by other observables. Lastly, we present a UV complete model capable of generating dim-8 operators at the leading order, thus offering a unique perspective on the interplay between different operator dimensions in probing new physics phenomena., Comment: 20 pages, 4 figures

The effectiveness of large language models (LLMs) is often hindered by duplicated data in their extensive pre-training datasets. Current approaches primarily focus on detecting and removing duplicates, which risks the loss of valuable information and neglects the varying degrees of duplication. To address this, we propose a soft deduplication method that maintains dataset integrity while selectively reducing the sampling weight of data with high commonness. Central to our approach is the concept of "data commonness", a metric we introduce to quantify the degree of duplication by measuring the occurrence probabilities of samples using an n-gram model. Empirical analysis shows that this method significantly improves training efficiency, achieving comparable perplexity scores with at least a 26% reduction in required training steps. Additionally, it enhances average few-shot downstream accuracy by 1.77% when trained for an equivalent duration. Importantly, this approach consistently improves performance, even on rigorously deduplicated datasets, indicating its potential to complement existing methods and become a standard pre-training process for LLMs., Comment: 12 pages, 7 figures

In this paper, we systematically investigate the general dark matter-electron interactions within the framework of effective field theories (EFT). We consider both the non-relativistic (NR) EFT and the relativistic EFT descriptions of the interactions with the spin of dark matter (DM) up to one, i.e., the scalar ($\phi$), fermion ($\chi$), and vector $(X)$ DM scenarios. We first collect the leading-order NR EFT operators describing the DM-electron interactions, and construct especially the NR operators for the vector DM case. Next, we consider all possible leading-order relativistic EFT operators including those with a photon field and perform the NR reduction to match them onto the NR EFT. Then we rederive the DM-bound-electron scattering rate within the NR EFT framework and find that the matrix element squared, which is the key input that encodes the DM and atomic information, can be compactly decomposed into three terms. Each term is a product of a DM response function $(a_{0,1,2})$, which is essentially a factor of Wilson coefficients squared, and its corresponding generalized atomic response function ($\widetilde W_{0,1,2}$). Lastly, we employ the electron recoil data from the DM direct detection experiments (including XENON10, XENON1T, and PandaX-4T) to constrain all the non-relativistic and relativistic operators in all three DM scenarios. We set strong bounds on the DM-electron interactions in the sub-GeV region. Particularly, we find that the latest PandaX-4T S2-only data provide stringent constraints on dark matter with a mass greater than approximately 20 MeV, surpassing those from the previous XENON10 and XENON1T experiments., Comment: 37 pages, 8 figures, minor corrections for several typos, and added one reference. Accepted by JHEP

In this Letter we revisit general dark matter (DM)-bound-electron interactions studied previously in the influential work [R. Catena {\it et al.,} Atomic responses to general dark matter-electron interactions, Phys. Rev. Res. 2, 033195 (2020)] For the most general DM-electron nonrelativistic or relativistic interactions for DM with spin up to 1, we find the average ionization matrix element squared can be organized into three terms, each of which is a product of a DM response function ($a_{0,1,2}$) and a linear combination ($\widetilde W_{0,1,2}$) of the four atomic response functions ($W_{1,2,3,4}$) given in that work, $ \widetilde W_0 = W_1, \, \widetilde W_1 = |\pmb{v}_0^\perp|^2 W_1 - 2 {m_e\, \pmb{q}\cdot \pmb{v}_0^\perp \over \pmb{q}^2} W_2 + W_3,\, \widetilde W_2 = { (\pmb{q}\cdot \pmb{v}_0^\perp)^2 \over \pmb{q}^2} W_1 - 2 {m_e\, \pmb{q}\cdot \pmb{v}_0^\perp \over \pmb{q}^2} W_2 + {m_e^2 \over \pmb{q}^2}W_4$. Furthermore, we find a crucial minus sign was missed for the calculation of $W_2$ in that work, which has significant phenomenological consequences when explaining experimental bounds on specific DM scenarios. Due to the corrected sign, there can be significant cancellations between the $W_2$ and $W_{3,4}$ terms, so that $\widetilde W_{1,2}$ are dominated by the usual response function $W_1$ in some cases. Many DM scenarios involving DM or electron axial-vector current can yield $W_2$ and thus are potentially affected by the sign. As an example, we show that the recent XENON1T constraint on the fermionic DM anapole moment is weakened by a factor of 2 or so. We also present a complete list of NR operators for spin-1 DM and compute their contributions to the DM response functions., Comment: v2: 7+2 pages, 4 figures, matches the published version as a letter in PRD

Understanding the interactions of atoms such as forces in 3D atomistic systems is fundamental to many applications like molecular dynamics and catalyst design. However, simulating these interactions requires compute-intensive ab initio calculations and thus results in limited data for training neural networks. In this paper, we propose to use denoising non-equilibrium structures (DeNS) as an auxiliary task to better leverage training data and improve performance. For training with DeNS, we first corrupt a 3D structure by adding noise to its 3D coordinates and then predict the noise. Different from previous works on denoising, which are limited to equilibrium structures, the proposed method generalizes denoising to a much larger set of non-equilibrium structures. The main difference is that a non-equilibrium structure does not correspond to local energy minima and has non-zero forces, and therefore it can have many possible atomic positions compared to an equilibrium structure. This makes denoising non-equilibrium structures an ill-posed problem since the target of denoising is not uniquely defined. Our key insight is to additionally encode the forces of the original non-equilibrium structure to specify which non-equilibrium structure we are denoising. Concretely, given a corrupted non-equilibrium structure and the forces of the original one, we predict the non-equilibrium structure satisfying the input forces instead of any arbitrary structures. Since DeNS requires encoding forces, DeNS favors equivariant networks, which can easily incorporate forces and other higher-order tensors in node embeddings. We study the effectiveness of training equivariant networks with DeNS on OC20, OC22 and MD17 datasets and demonstrate that DeNS can achieve new state-of-the-art results on OC20 and OC22 and significantly improve training efficiency on MD17., Comment: Published in Transactions on Machine Learning Research (TMLR)

Effective field theory (EFT) provides a model-independent framework for interpreting the results of dark matter (DM) direct detection experiments. In this study, we demonstrate that the two fermionic DM-quark tensor operators, $(\bar{\chi} i\sigma^{\mu\nu} \gamma^5 \chi) (\bar{q} \sigma_{\mu\nu}q)$ and $(\bar{\chi} \sigma^{\mu\nu} \chi) (\bar{q} \sigma_{\mu\nu} q)$, can contribute to the DM electric and magnetic dipole moments via nonperturbative QCD effects, in addition to the well-studied contact DM-nucleon operators. We then investigate the constraints on these two operators by considering both the contact and the dipole contributions using the XENON1T nuclear recoil and Migdal effect data. We also recast other existing bounds on the DM dipole operators, derived from electron and nuclear recoil measurements in various direct detection experiments, as constraints on the two tensor operators. For $m_\chi \lesssim 1\,\rm GeV$, our results significantly extend the reach of constraints on the DM-quark tensor operators to masses as low as $5\,\rm MeV$, with the bound exceeding that obtained by the Migdal effect with only contact interactions by an order of magnitude or so. In particular, for the operator $(\bar{\chi} \sigma^{\mu\nu}i\gamma_5 \chi) (\bar{q} \sigma_{\mu\nu}q)$ with DM mass $m_\chi \gtrsim 10\,\rm GeV$, the latest PandaX constraint on the DM electric dipole moment puts more stringent bounds than the previous direct detection limit., Comment: v1: 19 pages, 4 figures; v2: We have updated discussion on nucleon form factors by using more recent numbers, without modifying our final numerical results; v3: Add comparisons with other constraints and provide a UV model for the generation of tensor operators from tree-level matching; To be published in CPC

With the method of the background field expansion, we investigate the one-loop quantization of the Euclidean nonlocal $f(R)$ model in the de-Sitter universe. We obtain the ghost-free condition (GFC) based on the transformation from the Jordan frame to the Einstein frame and the classical stability condition (CSC) satisfied $f^{(0)}_{RR}-\phi_0F^{(0)}_{RR}<0$. We present the on-shell and off-shell one-loop effective action and quantum stability condition (QSC) by utilizing the generalized zeta-function. We find that under the fulfillment of GFC, CSC and QSC are inconsistent., Comment: 26 pages,0 figure

We introduce the effective field theory of two different light dark particles interacting with the standard model (SM) light states in a single vertex, termed dark sector effective field theory (DSEFT). We focus on the new light particles with spin up to 1 and being real in essence, namely, new real scalars $\phi$ and $S$, Majorana fermions $\chi$ and $\psi$, and real vectors $X_\mu$ and $V_\mu$. In the framework of low energy effective field theory with QED and QCD symmetry, the DSEFT can be classified into six categories, including the scalar-scalar-SM ($\phi S$-SM), fermion-fermion-SM ($\chi\psi$-SM), vector-vector-SM ($X V$-SM), scalar-fermion-SM ($\phi \chi$-SM), scalar-vector-SM ($\phi X$-SM), and fermion-vector-SM ($\chi X$-SM) cases. For each case, we construct the effective operator basis up to canonical dimension 7, which will cover most interesting phenomenology at low energy. As a phenomenological example, we investigate the longstanding neutron lifetime anomaly through the neutron dark decay modes $n \to \chi \phi \text{ or } \chi X$ from the effective interactions in the fermion-scalar-SM or fermion-vector-SM case. When treating the light fermion as a dark matter candidate, we also explore the constraints from DM-neutron annihilation signal at Super-Kamiokande. We find the neutron dark decay in each scenario can accommodate the anomaly, at the same time, without contradicting with the Super-Kamiokande limit., Comment: 33 pages, 4 figures, typos are corrected and several new references are included. To appear in JHEP

Decisions made by convolutional neural networks(CNN) can be understood and explained by visualizing discriminative regions on images. To this end, Class Activation Map (CAM) based methods were proposed as powerful interpretation tools, making the prediction of deep learning models more explainable, transparent, and trustworthy. However, all the CAM-based methods (e.g., CAM, Grad-CAM, and Relevance-CAM) can only be used for interpreting CNN models with fully-connected (FC) layers as a classifier. It is worth noting that many deep learning models classify images without FC layers, e.g., few-shot learning image classification, contrastive learning image classification, and image retrieval tasks. In this work, a post-hoc interpretation tool named feature activation map (FAM) is proposed, which can interpret deep learning models without FC layers as a classifier. In the proposed FAM algorithm, the channel-wise contribution weights are derived from the similarity scores between two image embeddings. The activation maps are linearly combined with the corresponding normalized contribution weights, forming the explanation map for visualization. The quantitative and qualitative experiments conducted on ten deep learning models for few-shot image classification, contrastive learning image classification and image retrieval tasks demonstrate the effectiveness of the proposed FAM algorithm., Comment: 14 pages

For one-dimension Brownian motion in the confined system with the size $L$, the mean-squared displacement(MSD) defined by $\left \langle (x-x_0)^2 \right\rangle$ should be proportional to $t^{\alpha(t)}$. The power $\alpha(t)$ should range from $1$ to $0$ over time, and the MSD turns from $2Dt$ to $c L^2$, here the coefficient $c$ independent of $t$, $D$ being the diffusion coefficient. The paper aims to quantitatively solve the MSD in the intermediate confinement regime. The key to this problem is how to deal with the propagator and the normalization factor of the Fokker-Planck equation(FPE) with the Dirichlet Boundaries. Applying the Euler-Maclaurin approximation(EMA) and integration by parts for the small $t$, we obtain the MSD being $2Dt(1-\frac{2\sqrt{\xi} }{3\pi\sqrt{\pi}})$, with $t_{ch}=\frac{L^2}{4\pi^2D},\xi\equiv \frac{t}{t_{ch}}$, and the power $\alpha(t)$ being $\frac{1-0.18\sqrt{\xi}}{1-0.12\sqrt{\xi}}$. Further, we analysis the MSD and the power for the $d$-dimension system with $\gamma$-dimension confinement. In the case of $\gamma< d$, there exists the sub-diffusive behavior in the intermediate time. The universal description is consistent with the recent experiments and simulations in the micro-nano systems. Finally, we calculate the position variance(PV) meaning $\left\langle (x-\left\langle x \right\rangle)^2 \right\rangle$. Under the initial condition referring to the different probability density function(PDF) being $p_{0}(x)$, MSD and PV should exhibit different dependencies on time, which reflect corresponding diffusion behaviors.As examples, the paper discusses the representative initial PDFs reading $p_{0}(x)=\delta(x-x_0)$, with the midpoint $x_0=\frac{L}{2}$ and the endpoint $x_0=\epsilon$(or $0^+$).The MSD(equal to PV) reads $2Dt(1-\frac{5\pi^3 Dt}{L^2})$,and $\frac{4}{\pi}(2Dt)[1+\frac{2\sqrt{\pi Dt}}{L}]$for the small $t$,respectively., Comment: 18 pages, 4 figurea