SearcharxivSearch

arXiv subjects

R. J. Chen

Publications and source records attributed to R. J. Chen.

At least 19 recordsLinked to original sources

DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learning

General reasoning represents a long-standing and formidable challenge in artificial intelligence. Recent breakthroughs, exemplified by large language models (LLMs) and chain-of-thought prompting, have achieved considerable success on foundational reasoning tasks. However, this success is heavily contingent upon extensive human-annotated demonstrations, and models' capabilities are still insufficient for more complex problems. Here we show that the reasoning abilities of LLMs can be incentivized through pure reinforcement learning (RL), obviating the need for human-labeled reasoning trajectories. The proposed RL framework facilitates the emergent development of advanced reasoning patterns, such as self-reflection, verification, and dynamic strategy adaptation. Consequently, the trained model achieves superior performance on verifiable tasks such as mathematics, coding competitions, and STEM fields, surpassing its counterparts trained via conventional supervised learning on human demonstrations. Moreover, the emergent reasoning patterns exhibited by these large-scale models can be systematically harnessed to guide and enhance the reasoning capabilities of smaller models.

cs.CL

Bayesian and Monte Carlo approaches to estimating uncertainty for the measurement of the bound-state $β$ decay of $^{205}\mathrm{Tl}^{81+}$

The measurement of the bound-state $β$ decay of $^{205}\mathrm{Tl}^{81+}$ at the Experimental Storage Ring at GSI, Darmstadt, has recently been reported with substantial impact on the use of $^{205}\mathrm{Pb}$ as an early Solar System chronometer and the low-energy measurement of the solar neutrino spectrum via the LOREX project. Due to the technical challenges in producing a high-purity $^{205}\mathrm{Tl}^{81+}$ secondary beam, a robust statistical method needed to be developed to estimate the variation in the contaminant $^{205}\mathrm{Pb}^{81+}$ produced in the fragmentation reaction, which was subsequently transmitted and stored in the ESR. Here we show that Bayesian and Monte Carlo methods produced comparable estimates for the contaminant variation, each with unique advantages and challenges given the complex statistical problems for this experiment. We recommend the adoption of such methods in future experiments that exhibit unknown statistical fluctuations.

nucl-ex

Half-life determination of heavy ions in a storage ring considering feeding and depleting background processes

Heavy-ion storage rings have relatively large momentum acceptance which allows for multiple ion species to circulate at the same time. This needs to be considered in radioactive decay measurements of highly charged ions, where atomic charge exchange reactions can significantly alter the intensities of parent and daughter ions. In this study, we investigate this effect using the decay curves of ion numbers in the recent $^{205}$Tl$^{81+}$ bound-state beta decay experiment conducted using the Experimental Storage Ring at GSI Darmstadt. To understand the intricate dynamics of ion numbers, we present a set of differential equations that account for various atomic and nuclear reaction processes-bound-state beta decay, atomic electron recombination and capture, and electron ionization. By incorporating appropriate boundary conditions, we develop a set of differential equations that accurately simulate the decay curves of various simultaneously stored ions in the storage ring: $^{205}$Tl$^{81+}$, $^{205}$Pb$^{81+}$, $^{205}$Pb$^{82+}$, $^{200}$Hg$^{79+}$, and $^{200}$Hg$^{80+}$. Through a quantitative comparison between simulations and experimental data, we provide insights into the detailed reaction mechanisms governing stored heavy ions within the storage ring. Our approach effectively models charge-changing processes, reduces the complexity of the experimental setup, and provides a simpler method for measuring the decay half-lives of highly charged ions in storage rings.

nucl-ex

$Z=14$ Magicity Revealed by the Mass of the Proton Dripline Nucleus $^{22}$Si

Using the $Bρ$-defined isochronous mass spectrometry technique, we conducted the first mass measurement of the proton dripline nucleus $^{22}$Si. We confirm that $^{22}$Si is bound against particle emission with $S_p/S_{2p}=+1412(114)/+229(54)$ keV, fixing the proton dripline location for the Si element. By analyzing the mass differences of the neighboring $sd$-shell nuclei, we find that $^{22}$Si exhibits a doubly-magic character similar to its mirror partner $^{22}$O, and that the mirror energy difference of $^{22}$Si-$^{22}$O deviates from the predictions assuming mirror symmetry. Gamow shell-model calculations reveal that the average occupations of valence protons in $^{22}$Si are nearly identical to those of valence neutrons in $^{22}$O, supporting the $Z=14$ magicity in $^{22}$Si. The observed mirror-symmetry breaking is attributed to the extended proton distribution in $^{22}$Si arising from a small contribution of the unbound $\pi2s_{1/2}$ orbital.

nucl-ex

DeepSeek-V3 Technical Report

We present DeepSeek-V3, a strong Mixture-of-Experts (MoE) language model with 671B total parameters with 37B activated for each token. To achieve efficient inference and cost-effective training, DeepSeek-V3 adopts Multi-head Latent Attention (MLA) and DeepSeekMoE architectures, which were thoroughly validated in DeepSeek-V2. Furthermore, DeepSeek-V3 pioneers an auxiliary-loss-free strategy for load balancing and sets a multi-token prediction training objective for stronger performance. We pre-train DeepSeek-V3 on 14.8 trillion diverse and high-quality tokens, followed by Supervised Fine-Tuning and Reinforcement Learning stages to fully harness its capabilities. Comprehensive evaluations reveal that DeepSeek-V3 outperforms other open-source models and achieves performance comparable to leading closed-source models. Despite its excellent performance, DeepSeek-V3 requires only 2.788M H800 GPU hours for its full training. In addition, its training process is remarkably stable. Throughout the entire training process, we did not experience any irrecoverable loss spikes or perform any rollbacks. The model checkpoints are available at https://github.com/deepseek-ai/DeepSeek-V3.

cs.CL

Bound-State Beta Decay of $\mathbf{\mathrm{^{205}{Tl}^{81+}}}$ Ions and the LOREX Project

Stable $^{205}$Tl ions have the lowest known energy threshold for capturing electron neutrinos ($ν_e$) of ${ E}_{ν_e}\ge50.6$\,keV. The Lorandite Experiment (LOREX), proposed in the 1980s, aims at obtaining the longtime averaged solar neutrino flux by utilizing natural deposits of Tl-bearing lorandite ores. To determine the $ν_e$ capture cross section, it is required to know the strength of the weak transition connecting the ground state of $^{205}$Tl and the 2.3 keV first excited state in $^{205}$Pb. The only way to experimentally address this transition is to measure the bound-state beta decay ($β_{b}$) of fully ionized $\mathrm{^{205}Tl^{81+}}$ ions. After three decades of meticulous preparation, the half-life of the $β_{b}$ decay of $\mathrm{^{205}Tl^{81+}}$ has been measured to be $291_{-27}^{+33}$ days using the Experimental Storage Ring (ESR) at GSI, Darmstadt. The longer measured half-life compared to theoretical estimates reduces the expected signal-to-noise ratio in the LOREX, thus challenging its feasibility.

nucl-ex

High-temperature $^{205}$Tl decay clarifies $^{205}$Pb dating in early Solar System

Radioactive nuclei with lifetimes on the order of millions of years can reveal the formation history of the Sun and active nucleosynthesis occurring at the time and place of its birth. Among such nuclei whose decay signatures are found in the oldest meteorites, $^{205}$Pb is a powerful example, as it is produced exclusively by slow neutron captures (the s process), with most being synthesized in asymptotic giant branch (AGB) stars. However, making accurate abundance predictions for $^{205}$Pb has so far been impossible because the weak decay rates of $^{205}$Pb and $^{205}$Tl are very uncertain at stellar temperatures. To constrain these decay rates, we measured for the first time the bound-state $β^-$ decay of fully ionized $^{205}$Tl$^{81+}$, an exotic decay mode that only occurs in highly charged ions. The measured half-life is 4.7 times longer than the previous theoretical estimate and our 10% experimental uncertainty has eliminated the main nuclear-physics limitation. With new, experimentally backed decay rates, we used AGB stellar models to calculate $^{205}$Pb yields. Propagating those yields with basic galactic chemical evolution (GCE) and comparing with the $^{205}$Pb/$^{204}$Pb ratio from meteorites, we determined the isolation time of solar material inside its parent molecular cloud. We find positive isolation times that are consistent with the other s-process short-lived radioactive nuclei found in the early Solar System. Our results reaffirm the site of the Sun's birth as a long-lived, giant molecular cloud and support the use of the $^{205}$Pb--$^{205}$Tl decay system as a chronometer in the early Solar System.

nucl-ex

Nuclear structure of dripline nuclei elucidated through precision mass measurements of $^{23}$Si, $^{26}$P, $^{27,28}$S, and $^{31}$Ar

Using the B$ρ$-defined isochronous mass spectrometry technique, we report the first determination of the $^{23}$Si, $^{26}$P, $^{27}$S, and $^{31}$Ar masses and improve the precision of the $^{28}$S mass by a factor of 11. Our measurements confirm that these isotopes are bound and fix the location of the proton dripline in P, S, and Ar. We find that the mirror energy differences of the mirror-nuclei pairs $^{26}$P-$^{26}$Na, $^{27}$P-$^{27}$Mg, $^{27}$S-$^{27}$Na, $^{28}$S-$^{28}$Mg, and $^{31}$Ar-$^{31}$Al deviate significantly from the values predicted assuming mirror symmetry. In addition, we observe similar anomalies in the excited states, but not in the ground states, of the mirror-nuclei pairs $^{22}$Al-$^{22}$F and $^{23}$Al-$^{23}$Ne. Using $ab~ initio$ VS-IMSRG and mean field calculations, we show that such a mirror-symmetry breaking phenomeon can be explained by the extended charge distributions of weakly-bound, proton-rich nuclei. When observed, this phenomenon serves as a unique signature that can be valuable for identifying proton-halo candidates.

nucl-ex

DeepSeek-V2: A Strong, Economical, and Efficient Mixture-of-Experts Language Model

We present DeepSeek-V2, a strong Mixture-of-Experts (MoE) language model characterized by economical training and efficient inference. It comprises 236B total parameters, of which 21B are activated for each token, and supports a context length of 128K tokens. DeepSeek-V2 adopts innovative architectures including Multi-head Latent Attention (MLA) and DeepSeekMoE. MLA guarantees efficient inference through significantly compressing the Key-Value (KV) cache into a latent vector, while DeepSeekMoE enables training strong models at an economical cost through sparse computation. Compared with DeepSeek 67B, DeepSeek-V2 achieves significantly stronger performance, and meanwhile saves 42.5% of training costs, reduces the KV cache by 93.3%, and boosts the maximum generation throughput to 5.76 times. We pretrain DeepSeek-V2 on a high-quality and multi-source corpus consisting of 8.1T tokens, and further perform Supervised Fine-Tuning (SFT) and Reinforcement Learning (RL) to fully unlock its potential. Evaluation results show that, even with only 21B activated parameters, DeepSeek-V2 and its chat versions still achieve top-tier performance among open-source models.

cs.CL

Ground-state mass of $^{22}$Al and test of state-of-the-art \textit{ab initio} calculations

The ground-state mass excess of the $T_{z}=-2$ drip-line nucleus $^{22}$Al is measured for the first time to be $18103(10)$ keV using the newly-developed B$ρ$-defined isochronous mass spectrometry method at the cooler storage ring in Lanzhou. The new mass excess value allowed us to determine the excitation energies of the two low-lying $1^+$ states in $^{22}$Al with significantly reduced uncertainties of 51 keV. Comparing to the analogue states in its mirror nucleus $^{22}$F, the mirror energy differences of the two $1^+$ states in the $^{22}$Al-$^{22}$F mirror pair are determined to be $-625(51)$ keV and $-330(51)$ keV, respectively. The excitation energies and the mirror energy differences are used to test the state-of-the-art \textit{ab initio} valence-space in-medium similarity renormalization group calculations with four sets of interactions derived from the chiral effective field theory. The mechanism leading to the large mirror energy differences is investigated and attributed to the occupation of the $πs_{1/2}$ orbital.

nucl-ex

Measurement of the Isolated Nuclear Two-Photon Decay in $^{72}\mathrm{Ge}$

The nuclear two-photon or double-gamma ($2γ$) decay is a second-order electromagnetic process whereby a nucleus in an excited state emits two gamma rays simultaneously. To be able to directly measure the $2γ$ decay rate in the low-energy regime below the electron-positron pair-creation threshold, we combined the isochronous mode of a storage ring with Schottky resonant cavities. The newly developed technique can be applied to isomers with excitation energies down to $\sim100$\,keV and half-lives as short as $\sim10$\,ms. The half-life for the $2γ$ decay of the first-excited $0^+$ state in bare $^{72}\mathrm{Ge}$ ions was determined to be $23.9\left(6\right)$\,ms, which strongly deviates from expectations.

nucl-ex

$Bρ$-defined isochronous mass spectrometry: a new approach for high-precision mass measurements of short-lived nuclei

A novel technique for broadband high-precision mass measurements of short-lived exotic nuclides is reported. It is based on the isochronous mass spectrometry (IMS) and realizes simultaneous determinations of revolution time and velocity of short-lived stored ions at the cooler storage ring CSRe in Lanzhou. The new technique, named as the $Bρ$-defined IMS or $Bρ$-IMS, boosts the efficiency, sensitivity, and accuracy of mass measurements, and is applied here to measure masses of neutron-deficient $fp$-shell nuclides. In a single accelerator setting, masses of $^{46}$Cr, $^{50}$Fe and $^{54}$Ni are determined with relative uncertainties of (5~-~6)$\times10^{-8}$, thereby improving the input data for testing the unitarity of the Cabibbo-Kobayashi-Maskawa quark mixing matrix. This is the technique of choice for future high-precision measurements of the most rarely produced shortest-lived nuclides.

physics.ins-det

$Bρ$-defined Isochronous Mass Spectrometry and Mass Measurements of $^{58}$Ni Fragments

A novel isochronous mass spectrometry, termed as $Bρ$-defined IMS, is established at the experimental cooler-storage ring CSRe in Lanzhou. Its potential has been studied through high precision mass measurements of $^{58}$Ni projectile fragments. Two time-of-flight detectors were installed in one of the straight sections of CSRe, thus enabling simultaneous measurements of the velocity and the revolution time of each stored short-lived ion. This allows for calculating the magnetic rigidity $Bρ$ and the orbit length $C$ of each ion. The accurate $Bρ(C)$ function has been constructed, which is a universal calibration curve used to deduce the masses of the stored nuclides. The sensitivity to single stored ions, quickness, and background-free characteristics of the method are ideally suited to address nuclides with very short lifetimes and tiniest production yields. In the limiting case of just a single particle, the attained mass resolving power allows one us to determine its mass-over-charge ratio $m/q$ with a remarkable precision of merely $\sim5$ keV. Masses of $T_z = -3/2$ fp-shell nuclides are re-determined with high accuracy, and the validity of the isospin multiplet mass equation is tested up to the heaviest isospin quartet with $A = 55$. The new masses are also used to investigate the mirror symmetry of empirical residual proton-neutron interactions.

nucl-ex

Single and double $K$-shell vacancy production in slow Xe$^\textrm{54+,53+}$-Xe collisions

We present an experimental and theoretical study of symmetric $\textrm{Xe}^{54+}+\textrm{Xe}$ collisions at 50, 30, and 15 MeV/u, corresponding to strong perturbations with $v_K/v_\text{p}$ = 1.20, 1.55, and 2.20, respectively ($v_K$: classical $K$-shell orbital velocity, $v_\text{p}$: projectile velocity), as well as $\textrm{Xe}^{53+}+\textrm{Xe}$ collisions at 15 MeV/u. For each of these systems, x-ray spectra were measured under a forward angle of $35^\circ$ with respect to the projectile beam. Target satellite and hypersatellite radiation, $Kα_{2,1}^\mathrm{s}$ and $Kα_{2,1}^\mathrm{hs}$, respectively, were analyzed and used to derive cross section ratios for double-to-single target $K$-shell vacancy production. We compare our experimental results to relativistic time-dependent two-center calculations.

physics.atom-ph

Particle identification and revolution time corrections for the isochronous mass spectrometry in storage rings

In the isochronous mass spectrometry (IMS) performed at storage rings, masses of short-lived nuclides are determined through precision measurements of their mean revolution times. However, the distribution of revolution times could be seriously deteriorated by instabilities of the ring's magnetic fields. This becomes a significant obstacle for the particle identifications and mass determinations. A data analysis method is described in this paper which is able to largely remove the uncertainties caused by the magnetic field instabilities in the particle identifications and the mean revolution times. We show that this method is very effective for the IMS experiments even when the magnetic fields of a storage ring vary slowly up to a level of $ΔB/B\sim 10^{-4}$.

nucl-ex

Mass measurements for $T_{z}=-2$ $fp$-shell nuclei $^{40}$Ti, $^{44}$Cr, $^{46}$Mn, $^{48}$Fe, $^{50}$Co and $^{52}$Ni

By using isochronous mass spectrometry (IMS) at the experimental cooler storage ring CSRe, masses of short-lived $^{44}$Cr, $^{46}$Mn, $^{48}$Fe, $^{50}$Co and $^{52}$Ni were measured for the first time and the precision of the mass of $^{40}$Ti was improved by a factor of about 2. Relative precisions of $δm/m=(1-2)\times$10$^{-6}$ have been achieved. Details of the measurements and data analysis are described. The obtained masses are compared with the Atomic-Mass Evaluation 2016 (AME$^{\prime}$16) and with theoretical model predictions. The new mass data enable us to extract the higher order coefficients, $d$ and $e$, of the quartic form of the isobaric multiplet mass equation (IMME) for the $fp$-shell isospin quintets. Unexpectedly large $d$- and $e$-values for $A=44$ quintet are found. By re-visiting the previous experimental data on $β$-delayed protons from $^{44}$Cr decay, it is suggested that the observed anomaly could be due to the misidentification of the $T=2$, $J^π=0^{+}$ isobaric analog state (IAS) in $^{44}$V.

nucl-ex

Determination of luminosity for in-ring reactions: A new approach for the low-energy domain

Luminosity is a measure of the colliding frequency between beam and target and it is a crucial parameter for the measurement of absolute values, such as reaction cross sections. In this paper, we make use of experimental data from the ESR storage ring to demonstrate that the luminosity can be precisely determined by modelling the measured Rutherford scattering distribution. The obtained results are in good agreement with an independent measurement based on the x-ray normalization method. Our new method provides an alternative way to precisely measure the luminosity in low-energy stored-beam configurations. This can be of great value in particular in dedicated low-energy storage rings where established methods are difficult or impossible to apply.

nucl-ex