Search

Abstract. Automatically generating scientific literature surveys is a valuable task that can significantly enhance research efficiency. However, the diverse and complex nature of information within a literature survey poses substantial challenges for generative models. In this paper, we design a series of prompts to systematically leverage large language models (LLMs), enabling the creation of comprehensive literature surveys through a step-by-step approach. Specifically, we design prompts to guide LLMs to sequentially generate the title, abstract, hierarchical headings, and the main content of the literature survey. We argue that this design enables the generation of the headings from a high-level perspective. During the content generation process, this design effectively harnesses relevant information while minimizing costs by restricting the length of both input and output content in LLM queries. Our implementation with Qwen-long achieved third place in the NLPCC 2024 Scientific Literature Survey Generation evaluation task, with an overall score only 0.03% lower than the second-place team. Additionally, our soft heading recall is 95.84%, the second best among the submissions. Thanks to the efficient prompt design and the low cost of the Qwen-long API, our method reduces the expense for generating each literature survey to 0.1 RMB, enhancing the practical value of our method., Comment: NLPCC 2024

Long-range dipole-dipole interactions in subwavelength arrays of quantum emitters involve virtual photon exchange processes that impart forces on the emitters due to the imposed photon recoil. We perform a semi-classical analysis of the dynamics of an impurity allowed to freely move through a subwavelength array of atoms in different parameter regimes. We numerically solve the coupled set of equations between motional and spin degrees of freedom to elucidate the possible impurity trajectories realizable in this system. We find that the impurity can maintain quasi-stable orbits within the plaquette for long times. The regions through which these orbits pass are strongly dependent on the chosen atomic transition dipole moment. We further provide intuition for our findings based on a simplified model, where the lattice dynamics is adiabatically eliminated. As a final point of analysis, we also take the motional degrees of freedom of the lattice atoms into account, and study the polaron-like excitation induced in the kinetic state of the lattice by the impurity., Comment: 11 pages, 7 figures

Generative recommendation has emerged as a promising paradigm aimed at augmenting recommender systems with recent advancements in generative artificial intelligence. This task has been formulated as a sequence-to-sequence generation process, wherein the input sequence encompasses data pertaining to the user's previously interacted items, and the output sequence denotes the generative identifier for the suggested item. However, existing generative recommendation approaches still encounter challenges in (i) effectively integrating user-item collaborative signals and item content information within a unified generative framework, and (ii) executing an efficient alignment between content information and collaborative signals. In this paper, we introduce content-based collaborative generation for recommender systems, denoted as ColaRec. To capture collaborative signals, the generative item identifiers are derived from a pretrained collaborative filtering model, while the user is represented through the aggregation of interacted items' content. Subsequently, the aggregated textual description of items is fed into a language model to encapsulate content information. This integration enables ColaRec to amalgamate collaborative signals and content information within an end-to-end framework. Regarding the alignment, we propose an item indexing task to facilitate the mapping between the content-based semantic space and the interaction-based collaborative space. Additionally, a contrastive loss is introduced to ensure that items with similar collaborative GIDs possess comparable content representations, thereby enhancing alignment. To validate the efficacy of ColaRec, we conduct experiments on three benchmark datasets. Empirical results substantiate the superior performance of ColaRec.

Sequential recommendation (SR) models are typically trained on user-item interactions which are affected by the system exposure bias, leading to the user preference learned from the biased SR model not being fully consistent with the true user preference. Exposure bias refers to the fact that user interactions are dependent upon the partial items exposed to the user. Existing debiasing methods do not make full use of the system exposure data and suffer from sub-optimal recommendation performance and high variance. In this paper, we propose to debias sequential recommenders through Distributionally Robust Optimization (DRO) over system exposure data. The key idea is to utilize DRO to optimize the worst-case error over an uncertainty set to safeguard the model against distributional discrepancy caused by the exposure bias. The main challenge to apply DRO for exposure debiasing in SR lies in how to construct the uncertainty set and avoid the overestimation of user preference on biased samples. Moreover, how to evaluate the debiasing effect on biased test set is also an open question. To this end, we first introduce an exposure simulator trained upon the system exposure data to calculate the exposure distribution, which is then regarded as the nominal distribution to construct the uncertainty set of DRO. Then, we introduce a penalty to items with high exposure probability to avoid the overestimation of user preference for biased samples. Finally, we design a debiased self-normalized inverse propensity score (SNIPS) evaluator for evaluating the debiasing effect on the biased offline test set. We conduct extensive experiments on two real-world datasets to verify the effectiveness of the proposed methods. Experimental results demonstrate the superior exposure debiasing performance of proposed methods. Codes and data are available at \url{https://github.com/nancheng58/DebiasedSR_DRO}., Comment: Accept by WSDM 2024

Learning reinforcement learning (RL)-based recommenders from historical user-item interaction sequences is vital to generate high-reward recommendations and improve long-term cumulative benefits. However, existing RL recommendation methods encounter difficulties (i) to estimate the value functions for states which are not contained in the offline training data, and (ii) to learn effective state representations from user implicit feedback due to the lack of contrastive signals. In this work, we propose contrastive state augmentations (CSA) for the training of RL-based recommender systems. To tackle the first issue, we propose four state augmentation strategies to enlarge the state space of the offline data. The proposed method improves the generalization capability of the recommender by making the RL agent visit the local state regions and ensuring the learned value functions are similar between the original and augmented states. For the second issue, we propose introducing contrastive signals between augmented states and the state randomly sampled from other sessions to improve the state representation learning further. To verify the effectiveness of the proposed CSA, we conduct extensive experiments on two publicly accessible datasets and one dataset collected from a real-life e-commerce platform. We also conduct experiments on a simulated environment as the online evaluation setting. Experimental results demonstrate that CSA can effectively improve recommendation performance.

Recently, single-site observables have been shown to be useful for the detection of dynamical criticality due to an emergence of a universal critically-prethermal temporal regime in the magnetization [arXiv:2105.05986]. Here, we explore the potential of single-site observables as probes of quantum phase transitions in integrable and nonintegrable transverse-field Ising chains (TFIC). We analytically prove the requirement of zero modes for a quasi-stationary temporal regime to emerge at a bulk probe site, and show how this regime gives rise to a non-analytic behavior in the dynamical order profiles. Our $t$-DMRG calculations verify the results of the quench mean-field theory for near-integrable TFIC both with finite-size and finite-time scaling analyses. We find that both finite-size and finite-time analyses suggest a dynamical critical point for a strongly nonintegrable and locally connected TFIC. We finally demonstrate the presence of a quasi-stationary regime in the power-law interacting TFIC, and extract local dynamical order profiles for TFIC in the long-range Ising universality class with algebraic light cones., Comment: 22 pages, 23 figures. Revised with new analytical calculations, new content and improved presentation

We reveal a prethermal temporal regime upon suddenly quenching to the vicinity of a quantum phase transition in the time evolution of 1D spin chains. The prethermal regime is analytically found to be self-similar, and its duration is governed by the ground-state energy gap. Based on analytical insights and numerical evidence, we show that this critically prethermal regime universally exists independently of the location of the probe site, the presence of weak interactions, or the initial state. Moreover, the resulting prethermal dynamics leads to an out-of-equilibrium scaling function of the order parameter in the vicinity of the transition., Comment: Revised with an analytical theory and improved presentation. Main text: 4 Pages, 2 Figures; Supplementary: 6 Pages, 5 Figures

Many synthetic quantum systems allow particles to have dispersion relations that are neither linear nor quadratic functions. Here, we explore single-particle scattering in general spatial dimension $D\geq 1$ when the density of states diverges at a specific energy. To illustrate the underlying principles in an experimentally relevant setting, we focus on waveguide quantum electrodynamics (QED) problems (i.e. $D=1$) with dispersion relation $\epsilon(k)=\pm |d|k^m$, where $m\geq 2$ is an integer. For a large class of these problems for any positive integer $m$, we rigorously prove that when there are no bright zero-energy eigenstates, the $S$-matrix evaluated at an energy $E\to 0$ converges to a universal limit that is only dependent on $m$. We also give a generalization of a key index theorem in quantum scattering theory known as Levinson's theorem -- which relates the scattering phases to the number of bound states -- to waveguide QED scattering for these more general dispersion relations. We then extend these results to general integer dimensions $D \geq 1$, dispersion relations $\epsilon(\boldsymbol{k}) = |\boldsymbol{k}|^a$ for a $D$-dimensional momentum vector $\boldsymbol{k}$ with any real positive $a$, and separable potential scattering.

Rydberg polaritons provide an example of a rare type of system where three-body interactions can be as strong or even stronger than two-body interactions. The three-body interactions can be either dispersive or dissipative, with both types possibly giving rise to exotic, strongly-interacting, and topological phases of matter. Despite past theoretical and experimental studies of the regime with dispersive interaction, the dissipative regime is still mostly unexplored. Using a renormalization group technique to solve the three-body Schr\"odinger equation, we show how the shape and strength of dissipative three-body forces can be universally enhanced for Rydberg polaritons. We demonstrate how these interactions relate to the transmission through a single-mode cavity, which can be used as a probe of the three-body physics in current experiments., Comment: 6+6 pages, 4 figures

We study the propagation of photons in a one-dimensional environment consisting of two non-interacting species of photons frustratingly coupled to a single spin-1/2. The ultrastrong frustrated coupling leads to an extreme mixing of the light and matter degrees of freedom, resulting in the disintegration of the spin and a breakdown of the "dressed-spin", or polaron, description. Using a combination of numerical and analytical methods, we show that the elastic response becomes increasingly weak at the effective spin frequency, showing instead an increasingly strong and broadband response at higher energies. We also show that the photons can decay into multiple photons of smaller energies. The total probability of these inelastic processes can be as large as the total elastic scattering rate, or half of the total scattering rate, which is as large as it can be. The frustrated spin induces strong anisotropic photon-photon interactions that are dominated by inter-species interactions. Our results are relevant to state-of-the-art circuit and cavity quantum electrodynamics experiments., Comment: 5+13 pages, 3 + 6 figures. v3: close to published version

Single photons coupled to atomic systems have shown to be a promising platform for developing quantum technologies. Yet a bright on-demand, highly pure and highly indistinguishable single-photon source compatible with atomic platforms is lacking. In this work, we demonstrate such a source based on a strongly interacting Rydberg system. The large optical nonlinearities in a blockaded Rydberg ensemble convert coherent light into a single-collective excitation that can be coherently retrieved as a quantum field. We observe a single-transverse-mode efficiency up to 0.18(2), $g^{(2)}=2.0(1.5)\times10^{-4}$, and indistinguishability of 0.982(7), making this system promising for scalable quantum information applications. Accounting for losses, we infer a generation probability up to 0.40(4). Furthermore, we investigate the effects of contaminant Rydberg excitations on the source efficiency. Finally, we introduce metrics to benchmark the performance of on-demand single-photon sources.

The quantum coupling of individual superconducting qubits to microwave photons leads to remarkable experimental opportunities. Here we consider the phononic case where the qubit is coupled to an electromagnetic surface acoustic wave antenna that enables supersonic propagation of the qubit oscillations. This can be considered as a giant atom that is many phonon wavelengths long. We study an exactly solvable toy model that captures these effects, and find that this non-Markovian giant atom has a suppressed relaxation, as well as an effective vacuum coupling between a qubit excitation and a localized wave packet of sound, even in the absence of a cavity for the sound waves. Finally, we consider practical implementations of these ideas in current surface acoustic wave devices., Comment: 8 pages, 4 figures