arXiv subjects

Chao Chen

Publications and source records attributed to Chao Chen.

At least 505 records · Page 28Linked to original sources

arXiv2024

Memorization in deep learning: A survey

Deep Learning (DL) powered by Deep Neural Networks (DNNs) has revolutionized various domains, yet understanding the intricacies of DNN decision-making and learning processes remains a significant challenge. Recent investigations have uncovered an interesting memorization phenomenon in which DNNs tend to memorize specific details from examples rather than learning general patterns, affecting model generalization, security, and privacy. This raises critical questions about the nature of generalization in DNNs and their susceptibility to security breaches. In this survey, we present a systematic framework to organize memorization definitions based on the generalization and security/privacy domains and summarize memorization evaluation methods at both the example and model levels. Through a comprehensive literature review, we explore DNN memorization behaviors and their impacts on security and privacy. We also introduce privacy vulnerabilities caused by memorization and the phenomenon of forgetting and explore its connection with memorization. Furthermore, we spotlight various applications leveraging memorization and forgetting mechanisms, including noisy label learning, privacy preservation, and model enhancement. This survey offers the first-in-kind understanding of memorization in DNNs, providing insights into its challenges and opportunities for enhancing AI development while addressing critical ethical concerns.

cs.LG↗

arXiv2024

Measurements of the branching fractions of the $P$-wave charmonium spin-singlet state $h_c(^1P_1) \to h^+ h^-π^0/η$

Based on $(2712.4\pm 14.3)\times10^{6}$ $ψ(3686)$ events, we investigate four hadronic decay modes of the $P$-wave charmonium spin-singlet state $h_c(^1P_1) \to h^+ h^- π^0/η$ ($h=π$ or $K$) via the process $ψ(3686) \to π^{0}h_c$ at BESIII. The $h_c \to π^+ π^- π^0$ decay is observed with a significance of 9.6$σ$ after taking into account systematic uncertainties. Evidences for $h_c \to K^+ K^- π^0$ and $h_c \to K^+ K^- η$ are found with significances of $3.5σ$ and $3.3σ$, respectively, after considering the systematic uncertainties. The branching fractions of these decays are measured to be $\mathcal{B}(h_c \to π^+ π^- π^0)=(1.36\pm0.16\pm0.14)\times10^{-3}$, $\mathcal{B}(h_c \to K^+ K^- π^0)=(3.26\pm0.84\pm0.36)\times10^{-4}$, and $\mathcal{B}(h_c \to K^+ K^- η)=(3.13\pm1.08\pm0.38)\times10^{-4}$, where the first uncertainties are statistical and the second are systematic. No significant signal of $h_c\toπ^+π^-η$ is found, and the upper limit of its decay branching fraction is determined to be $\mathcal{B}(h_c\toπ^+π^-η) < 4.0 \times 10^{-4}$ at 90% confidence level.

hep-ex↗

arXiv2024

Evidence of the Singly Cabibbo Suppressed decay $Λ_c^+\to pπ^0$

Evidence for the singly Cabibbo suppressed decay $Λ_c^+\to pπ^0$ is reported for the first time with a statistical significance of $3.7σ$ based on 6.0 $\mathrm{fb}^{-1}$ of $e^+e^-$ collision data collected at center-of-mass energies between 4.600 and 4.843 GeV with the BESIII detector at the BEPCII collider. The absolute branching fraction of $Λ_c^+\to pπ^0$ is measured to be $(1.56^{+0.72}_{-0.58}\pm0.20)\times 10^{-4}$. Combining with the branching fraction of $Λ_c^+\to nπ^+$, $(6.6\pm1.3)\times10^{-4}$, the ratio of the branching fractions of $Λ_c^+\to nπ^+$ and $Λ_c^+\to pπ^0$ is calculated to be $3.2^{+2.2}_{-1.2}$. As an important input for the theoretical models describing the decay mechanisms of charmed baryons, our result indicates that the non-factorizable contributions play an essential role and their interference with the factorizable contributions should not be significant. In addition, the absolute branching fraction of $Λ_c^+\to pη$ is measured to be $(1.63\pm0.31_{\rm stat}\pm0.11_{\rm syst}) \times10^{-3}$.

hep-ex↗

arXiv2024

An Iterative Associative Memory Model for Empathetic Response Generation

Empathetic response generation aims to comprehend the cognitive and emotional states in dialogue utterances and generate proper responses. Psychological theories posit that comprehending emotional and cognitive states necessitates iteratively capturing and understanding associated words across dialogue utterances. However, existing approaches regard dialogue utterances as either a long sequence or independent utterances for comprehension, which are prone to overlook the associated words between them. To address this issue, we propose an Iterative Associative Memory Model (IAMM) for empathetic response generation. Specifically, we employ a novel second-order interaction attention mechanism to iteratively capture vital associated words between dialogue utterances and situations, dialogue history, and a memory module (for storing associated words), thereby accurately and nuancedly comprehending the utterances. We conduct experiments on the Empathetic-Dialogue dataset. Both automatic and human evaluations validate the efficacy of the model. Variant experiments on LLMs also demonstrate that attending to associated words improves empathetic comprehension and expression.

cs.CL↗

arXiv2024

Study of the decays $χ_{cJ} \rightarrow Λ\barΛϕ$

Based on $(2712.4 \pm 14.3) \times 10^{6}$ $ e^{+}e^{-}\toψ(3686)$ events collected with the BESIII detector operating at the BEPCII collider, we report the first evidence of $χ_{c0}\to Λ\bar Λϕ$ decays and the first observation of $χ_{c1,2}\to Λ\bar Λϕ$ decays, with significances of $4.5σ$, $11.3σ$ and $13.0σ$, respectively. The decay branching fractions of $χ_{c0,1,2}\to Λ\bar Λϕ$ are measured to be $( 2.99\pm1.24\pm0.19) \times 10^{-5}$, $(6.01\pm0.90\pm0.40 )\times 10^{-5}$, and $(7.13\pm0.81\pm0.36) \times 10^{-5}$, where the first uncertainties are statistical and the second systematic. No obvious enhancement near the $Λ\barΛ$ production threshold or excited $Λ$ state is found in the $Λϕ$ (or $\barΛϕ$) system.

hep-ex↗

arXiv2024

Measurement of absolute branching fractions of $D_s^+$ hadronic decays

Using $e^+ e^-$ collision data collected at the BESIII detector at center-of-mass energies between 4.128 and 4.226 GeV, corresponding to an integrated luminosity of $7.33~{\rm fb}^{-1}$, we determine the absolute branching fractions of fifteen hadronic $D_s^{+}$ decays with a double-tag technique. In particular, we make precise measurements of the branching fractions $\mathcal{B}(D_s^+ \to K^+ K^- π^+)=(5.49 \pm 0.04 \pm 0.07)\%$, $\mathcal{B}(D_s^+ \to K_S^0 K^+)=(1.50 \pm 0.01 \pm 0.01)\%$ and $\mathcal{B}(D_s^+ \to K^+ K^- π^+ π^0)=(5.50 \pm 0.05 \pm 0.11)\%$, where the first uncertainties are statistical and the second ones are systematic. The \emph{CP} asymmetries in these decays are also measured and all are found to be compatible with zero.

hep-ex↗

arXiv2024

Large Language Model Watermark Stealing With Mixed Integer Programming

The Large Language Model (LLM) watermark is a newly emerging technique that shows promise in addressing concerns surrounding LLM copyright, monitoring AI-generated text, and preventing its misuse. The LLM watermark scheme commonly includes generating secret keys to partition the vocabulary into green and red lists, applying a perturbation to the logits of tokens in the green list to increase their sampling likelihood, thus facilitating watermark detection to identify AI-generated text if the proportion of green tokens exceeds a threshold. However, recent research indicates that watermarking methods using numerous keys are susceptible to removal attacks, such as token editing, synonym substitution, and paraphrasing, with robustness declining as the number of keys increases. Therefore, the state-of-the-art watermark schemes that employ fewer or single keys have been demonstrated to be more robust against text editing and paraphrasing. In this paper, we propose a novel green list stealing attack against the state-of-the-art LLM watermark scheme and systematically examine its vulnerability to this attack. We formalize the attack as a mixed integer programming problem with constraints. We evaluate our attack under a comprehensive threat model, including an extreme scenario where the attacker has no prior knowledge, lacks access to the watermark detector API, and possesses no information about the LLM's parameter settings or watermark injection/detection scheme. Extensive experiments on LLMs, such as OPT and LLaMA, demonstrate that our attack can successfully steal the green list and remove the watermark across all settings.

cs.CR↗

arXiv2024

ROPO: Robust Preference Optimization for Large Language Models

Preference alignment is pivotal for empowering large language models (LLMs) to generate helpful and harmless responses. However, the performance of preference alignment is highly sensitive to the prevalent noise in the preference data. Recent efforts for this problem either marginally alleviate the impact of noise without the ability to actually reduce its presence, or rely on costly teacher LLMs prone to reward misgeneralization. To address these challenges, we propose the RObust Preference Optimization (ROPO) framework, an iterative alignment approach that integrates noise-tolerance and filtering of noisy samples without the aid of external models. Specifically, ROPO iteratively solves a constrained optimization problem, where we dynamically assign a quality-aware weight for each sample and constrain the sum of the weights to the number of samples we intend to retain. For noise-tolerant training and effective noise identification, we derive a robust loss by suppressing the gradients of samples with high uncertainty. We demonstrate both empirically and theoretically that the derived loss is critical for distinguishing noisy samples from clean ones. Furthermore, inspired by our derived loss, we propose a robustness-guided rejection sampling technique to compensate for the potential important information in discarded queries. Experiments on three widely-used datasets with Mistral-7B and Llama-2-7B demonstrate that ROPO significantly outperforms existing preference alignment methods, with its superiority growing as the noise rate increases.

cs.LG↗

arXiv2024

Determination of the number of $ψ(3686)$ events taken at BESIII

The number of $ψ(3686)$ events collected by the BESIII detector during the 2021 run period is determined to be $(2259.3\pm 11.1)\times 10^6$ by counting inclusive $ψ(3686)$ hadronic events. The uncertainty is systematic and the statistical uncertainty is negligible. Meanwhile, the numbers of $ψ(3686)$ events collected during the 2009 and 2012 run periods are updated to be $(107.7\pm0.6)\times 10^6$ and $(345.4\pm 2.6)\times 10^6$, respectively. Both numbers are consistent with the previous measurements within one standard deviation. The total number of $ψ(3686)$ events in the three data samples is $(2712.4\pm14.3)\times10^6$.

hep-ex↗

arXiv2024

Study of the decays $χ_{cJ}\toΛ\barΛω$

Using $(27.12\pm 0.14)\times10^{8}$ $ψ(3686)$ events collected with the BESIII detector, we present the first observation of the decays $χ_{cJ}\toΛ\barΛω$, where $J=0, 1, 2$, with statistical significances of $11.7 σ, 11.2 σ$, and $11.8 σ$. The branching fractions of these decays are determined to be $\mathcal{B}(χ_{c0}\toΛ\barΛω)=({2.37 \pm 0.22 \pm 0.23}) \times 10^{-4}$, $\mathcal{B}(χ_{c1}\toΛ\barΛω)=({1.01 \pm 0.10 \pm 0.11}) \times 10^{-4}$, and $\mathcal{B}(χ_{c2}\toΛ\barΛω)=({1.40 \pm 0.13 \pm 0.17}) \times 10^{-4}$, where the first uncertainties are statistical and the second are systematic. We observe no clear intermediate structures.

hep-ex↗

arXiv2024

Precision measurement of the branching fraction of \boldmath $J/ψ\rightarrow K^+K^-$ via $ψ(2S)\rightarrow π^+π^-J/ψ$

Using a sample of $448.1 \times 10^6$ $ψ(2S)$ events collected with the BESIII detector, we perform a study of the decay $J/ψ\rightarrow K^+K^-$ via $ψ(2S)\rightarrow π^+π^-J/ψ$. The branching fraction of $J/ψ\rightarrow K^+K^-$ is determined to be $\mathcal{B}_{K^+K^-}=(3.072\pm 0.023({\rm stat.})\pm 0.050({\rm syst.}))\times 10^{-4}$, which is consistent with previous measurements but with significantly improved precision.

hep-ex↗

arXiv2024

SI-MIL: Taming Deep MIL for Self-Interpretability in Gigapixel Histopathology

Introducing interpretability and reasoning into Multiple Instance Learning (MIL) methods for Whole Slide Image (WSI) analysis is challenging, given the complexity of gigapixel slides. Traditionally, MIL interpretability is limited to identifying salient regions deemed pertinent for downstream tasks, offering little insight to the end-user (pathologist) regarding the rationale behind these selections. To address this, we propose Self-Interpretable MIL (SI-MIL), a method intrinsically designed for interpretability from the very outset. SI-MIL employs a deep MIL framework to guide an interpretable branch grounded on handcrafted pathological features, facilitating linear predictions. Beyond identifying salient regions, SI-MIL uniquely provides feature-level interpretations rooted in pathological insights for WSIs. Notably, SI-MIL, with its linear prediction constraints, challenges the prevalent myth of an inevitable trade-off between model interpretability and performance, demonstrating competitive results compared to state-of-the-art methods on WSI-level prediction tasks across three cancer types. In addition, we thoroughly benchmark the local and global-interpretability of SI-MIL in terms of statistical analysis, a domain expert study, and desiderata of interpretability, namely, user-friendliness and faithfulness.

cs.CV↗

arXiv2024

First study of antihyperon-nucleon scattering $\barΛp\rightarrow\barΛp$ and measurement of $Λp\rightarrowΛp$ cross section

Using $(10.087\pm0.044)\times10^{9}$ $J/ψ$ events collected with the BESIII detector at the BEPCII storage ring, the processes $Λp\rightarrowΛp$ and $\barΛp\rightarrow\barΛp$ are studied, where the $Λ/\barΛ$ baryons are produced in the process $J/ψ\rightarrowΛ\barΛ$ and the protons are the hydrogen nuclei in the cooling oil of the beam pipe. Clear signals are observed for the two reactions. The cross sections in $-0.9\leq\rm{cos}θ_{Λ/\barΛ}\leq0.9$ are measured to be $σ(Λp\rightarrowΛp)=(12.2\pm1.6_{\rm{stat}}\pm1.1_{\rm{sys}})$ mb and $σ(\barΛ p\rightarrow\barΛ p)=(17.5\pm2.1_{\rm{stat}}\pm1.6_{\rm{sys}})$ mb at the $Λ/\barΛ$ momentum of $1.074$ GeV/$c$ within a range of $\pm0.017$ GeV/$c$, where the $θ_{Λ/\barΛ}$ are the scattering angles of the $Λ/\barΛ$ in the $Λp/\barΛp$ rest frames. Furthermore, the differential cross sections of the two reactions are also measured, where there is a slight tendency of forward scattering for $Λp\rightarrowΛp$, and a strong forward peak for $\barΛp\rightarrow\barΛp$. We present an approach to extract the total elastic cross sections by extrapolation. The study of $\barΛp\rightarrow\barΛp$ represents the first study of antihyperon-nucleon scattering, and these new measurements will serve as important inputs for the theoretical understanding of the (anti)hyperon-nucleon interaction.

hep-ex↗

arXiv2024

Search for the leptonic decays $D^{+}\to e^+ν_e$ and $D^{+}\to μ^+ν_μ$

We present the first search for the leptonic decays $D^{*+}\to e^+ν_e$ and $D^{*+}\to μ^+ν_μ$ by analyzing a data sample of electron-positron collisions recorded with the BESIII detector at center-of-mass energies between 4.178 and 4.226 GeV, corresponding to an integrated luminosity of 6.32~fb$^{-1}$. No significant signal is observed. The upper limits on the branching fractions for $D^{*+}\to e^+ν_e$ and $D^{*+}\to μ^+ν_μ$ are set to be $1.1 \times 10^{-5}$ and $4.3 \times 10^{-6}$ at 90\% confidence level, respectively.

hep-ex↗

arXiv2024

First Observation of a Three-Resonance Structure in $e^+e^-\rightarrow$Nonopen Charm Hadrons

We report the measurement of the inclusive cross sections for $e^+e^-$$\rightarrow$nOCH (where nOCH denotes non-open charm hadrons) with improved precision at center-of-mass (c.m.) energies from 3.645 to 3.871 GeV. We observe three resonances: $\mathcal R(3760)$, $\mathcal R(3780)$, and $\mathcal R(3810)$ with significances of $8.1σ$, $13.7σ$, and $8.8σ$, respectively. The $\mathcal R(3810)$ state is observed for the first time, while the $\mathcal R(3760)$ and $\mathcal R(3780)$ states are observed for the first time in the nOCH cross sections. Two sets of resonance parameters describe the energy-dependent line shape of the cross sections well. In set I [set II], the $\mathcal R(3810)$ state has mass $(3805.7 \pm 1.1 \pm 2.7)$ [$(3805.7 \pm 1.1 \pm 2.7)$] MeV/$c^2$, total width $(11.6 \pm 2.9 \pm 1.9)$ [$(11.5 \pm 2.8 \pm 1.9)$] MeV, and an electronic width multiplied by the nOCH decay branching fraction of $(10.9\pm 3.8\pm 2.5)$ [$(11.0\pm 3.4\pm 2.5)$] eV. In addition, we measure the branching fractions ${\mathcal B}[{\mathcal R}(3760)$$\rightarrow$nOCH$]=(25.2 \pm 16.1 \pm 30.4)\% [(6.4 \pm 4.8 \pm 7.7)\%]$ and ${\mathcal B}[\mathcal R(3780)$$\rightarrow$nOCH$]=(12.3 \pm 6.6 \pm 8.3)\% [(10.4 \pm 4.8 \pm 7.0)\%]$ for the first time. The $\mathcal R(3760)$ state can be interpreted as an open-charm (OC) molecular state, but containing a simple four-quark state component. The $\mathcal R(3810)$ state can be interpreted as a hadrocharmonium state.

hep-ex↗

arXiv2024

Measurement of the ${e}^{+}{e}^{-}\to p \bar{p}π^{0}$ cross section at $\sqrt{s}=2.1000-3.0800$ GeV

The process $e^{+}e^{-}\to p\bar{p}π^{0}$ is studied at 20 center-of-mass energies ranging from 2.1000 to 3.0800 GeV using 636.8 pb$^{-1}$ of data collected with the BESIII detector operating at the BEPCII collider. The Born cross sections for $e^{+}e^{-}\to p\bar{p}π^{0}$ are measured with high precision. Since the lowest center-of-mass energy, 2.1000 GeV, is less than 90 MeV above the $p\bar{p}π^0$ energy threshold, we can probe the threshold behavior for this reaction. However, no anomalous threshold enhancement is found in the cross sections for $e^{+}e^{-}\to p\bar{p}π^{0}$.

hep-ex↗

arXiv2024

Binary Supermassive Black Holes Orbiting Dark Matter Solitons: From the Dual AGN in UGC4211 to NanoHertz Gravitational Waves

We explore orbital implications of the Supermassive Black Hole (SMBH) binary in UGC4211 for the energy spectrum of stochastic gravitational wave background (SGWB), measured with pulsar timing. The SMBH binary in UGC4211 has a projected separation of $\sim 230\,$pc and relative velocity of $\sim 150\,$km/s along the line of sight. It orbits with a disk of gas and stars, with a total mass $\sim 1.7 \times 10^9 M_\odot$ that is several times larger than the combined SMBHs plus the observed gas and stars. The unseen mass can be naturally explained by a soliton of wave dark matter present within the SMBH orbit. Such a scenario is encouraging as during galaxy merger, the two precursor galactic solitons are expected to combine to generate a new soliton and hence bind the two initial SMBHs efficiently. Generalizing this scenario to the cosmological population of SMBH binaries, we show the SGWB spectrum produced by late-stage inspiraling is modified preferentially at low frequency by the presence of soliton. Finally, we demonstrate that the NANOGrav and EPTA data can be well-fit in this scenario, favoring $\{m_a, f_a\} \sim \{10^{-21.7} {\rm eV}, 10^{15.5} {\rm GeV}\}$ and $\{10^{-20.5} {\rm eV}, 10^{16.8} {\rm GeV}\}$ respectively when the UGC4211 data and the constraints from dwarf galaxies are also combined.

astro-ph.HE↗

arXiv2024

Measurement of the absolute branching fraction of the three-body decay $Λ_{c}^+ \to Ξ^{0}K^{+}π^{0}$ and search for $Λ_{c}^+ \to nK^+π^0$, $Σ^{0}K^{+}π^{0}$ and $ΛK^{+}π^{0}$

The Cabbibo-favored decay $Λ_{c}^+ \to Ξ^{0}K^{+}π^{0}$ is studied for the first time using 6.1 fb$^{-1}$ of $e^+e^-$ collision data at center-of-mass energies between 4.600 and 4.840 GeV, collected with the BESIII detector at the BEPCII collider. With a double-tag method, the branching fraction of the three-body decay $Λ_{c}^+ \to Ξ^{0}K^{+}π^{0}$ is measured to be $(7.79 \pm 1.46 _{\rm} \pm0.71 _{\rm}) \times 10^{ - 3}$, where the first and second uncertainties are statistical and systematic, respectively. The branching fraction of the two-body decay $Λ_{c}^+ \to Ξ(1530)^{0}K^+$ is $(5.99\pm1.04\pm0.29)\times10^{-3}$, which is consistent with the previous result of $(5.02\pm0.99\pm0.31)\times 10^{-3}$. In addition, the upper limit on the branching fraction of the doubly Cabbibo-suppressed decay $Λ_{c}^+ \to nK^+π^0$ is $7.1 \times 10^{-4}$ at the 90$\%$ confidence level. The upper limits on the branching fractions of $Λ_{c}^+ \to Σ^{0}K^{+}π^{0}$ and $ΛK^{+}π^{0}$ are also determined to be $1.8\times 10^{-3}$ and $ 2.0 \times 10^{-3}$, respectively.

hep-ex↗