Searcharxiv⌕ Search

arXiv subjects

Yu Lan

Publications and source records attributed to Yu Lan.

At least 19 recordsLinked to original sources
arXiv2026

Episodic Memory Temporal Consistency for Cooperative Multi-Agent Reinforcement Learning

Cooperative Multi-Agent Reinforcement Learning (MARL) frequently suffers from severe reward sparsity and exploration bottlenecks. While episodic memory mechanisms mitigate these issues by reusing high-return trajectories, they often trap agents in local optima due to unconstrained incentive distribution and semantic representation collapse. To address this, we propose Episodic Memory Temporal Consistency (EMTC), a framework that robustly constructs and selectively leverages historical experiences. EMTC introduces two synergistic components: (1) a Temporally Consistent Semantic Embedder that integrates contrastive learning with time-conditioned state reconstruction, preventing representation collapse and enabling precise memory retrieval; and (2) a Temporal Consistency Gating Mechanism that dynamically modulates episodic incentives based on temporal consistency error. This adaptive gate filters misleading signals from pseudo-successful trajectories, effectively mitigating Q-value overestimation. We provide theoretical guarantees, establishing a strict error bound that directly links the observable temporal consistency error to the underlying trajectory optimality and representation quality. Extensive evaluations on the SMAC and GRF benchmarks demonstrate that EMTC consistently outperforms state-of-the-art baselines. Notably, compared to the strongest episodic baseline, EMTC achieves absolute win-rate improvements of up to 24% in super-hard SMAC scenarios and an average improvement of 28% across GRF tasks.

cs.LG↗
arXiv2025

LoongFlow: Directed Evolutionary Search via a Cognitive Plan-Execute-Summarize Paradigm

The transition from static Large Language Models (LLMs) to self-improving agents is hindered by the lack of structured reasoning in traditional evolutionary approaches. Existing methods often struggle with premature convergence and inefficient exploration in high-dimensional code spaces. To address these challenges, we introduce LoongFlow, a self-evolving agent framework that achieves state-of-the-art solution quality with significantly reduced computational costs. Unlike "blind" mutation operators, LoongFlow integrates LLMs into a cognitive "Plan-Execute-Summarize" (PES) paradigm, effectively mapping the evolutionary search to a reasoning-heavy process. To sustain long-term architectural coherence, we incorporate a hybrid evolutionary memory system. By synergizing Multi-Island models with MAP-Elites and adaptive Boltzmann selection, this system theoretically balances the exploration-exploitation trade-off, maintaining diverse behavioral niches to prevent optimization stagnation. We instantiate LoongFlow with a General Agent for algorithmic discovery and an ML Agent for pipeline optimization. Extensive evaluations on the AlphaEvolve benchmark and Kaggle competitions demonstrate that LoongFlow outperforms leading baselines (e.g., OpenEvolve, ShinkaEvolve) by up to 60% in evolutionary efficiency while discovering superior solutions. LoongFlow marks a substantial step forward in autonomous scientific discovery, enabling the generation of expert-level solutions with reduced computational overhead.

cs.AI↗
arXiv2025

Unusual electronic ordering in the pseudogap phase of underdoped cuprate superconductors

The pseudogap phase of the underdoped cuprate superconductors harbours diverse manifestations of different ordered electronic-states, and then these ordered electronic-states coexist or compete with superconductivity. Here starting from the microscopic electron propagator, the nature of the ordered electronic-states in the pseudogap phase is investigated within the $T$-matrix approach. This $T$-matrix is derived in terms of the inverse of matrix for various kinds of a single impurity, and then is used to evaluate the local density of states (LDOS) by the involvement of all the quasiparticle excitations and scattering processes. It is shown that a number of the anomalous properties in the underdoped cuprate superconductors is directly correlated to the opening of the normal-state pseudogap: (i) the structure of the microscopic octet scattering model generated by the normal-state pseudogap is essentially the same both in the superconducting (SC)-state and pseudogap phase, which naturally leads to that the quasiparticle scattering interference octet phenomenology observed in the SC-state exists in the pseudogap phase; (ii) however, the spectral weight at around the antinodal region in the SC-state is gapped out completely by both the SC gap and normal-state pseudogap, while it in the pseudogap phase is suppressed partially by the normal-state pseudogap, this directly leads to that the non-dispersive checkerboard charge ordering with a finite wave vector ${\bf Q}$ appears in the pseudogap phase only. The theory therefore also shows that the electronic-states affected by the normal-state pseudogap exhibit the LDOS modulation spectrum organization.

cond-mat.supr-con↗
arXiv2025

DEER: Disentangled Mixture of Experts with Instance-Adaptive Routing for Generalizable Machine-Generated Text Detection

Detecting machine-generated text has become a critical challenge amid the rapid advancement of LLMs, yet existing detectors degrade severely under domain shift. Through systematic pilot studies, we trace this vulnerability to two fundamental flaws in current generalization strategies, namely the incomplete preservation of domain-specific knowledge during multi-domain training and the misalignment between knowledge retrieval and the detection objective at inference. To address these gaps, we propose DEER, a Disentangled mixturE-of-ExpeRts framework that explicitly decouples domain-local and domain-invariant knowledge into specialized expert modules. Instead of static domain matching, DEER employs a reinforcement learning-driven router that selects expert pathways based on instance-level detection rewards. This task-aligned, domain-agnostic mechanism ensures robust adaptation to unseen distributions by prioritizing detection utility over stylistic resemblance. Extensive experiments demonstrate that DEER consistently outperforms state-of-the-art detectors, achieving average F1 improvements of 1.28% and 2.92%, and accuracy gains of 1.35% and 2.26% on in-domain and out-of-domain datasets, offering reliable generalization for open-world deployment.

cs.CL↗
arXiv2025

Can Reasoning Path still be Effective as Input? Bridging Post-Reasoning to Chain-of-Thought Compression

Recent developments have enabled advanced reasoning in Large Language Models (LLMs) via long Chain-of-Thought (CoT), trading efficiency during inference for performance. Existing works focus on compressing generated CoT in reasoning, which impairs the necessary information for deriving the correct answer. In this work, we propose post-reasoning, a reasoning paradigm that takes CoT as a part of context to simplify the reasoning task for LLMs. We find that post-reasoning significantly reduces the generation length of LLMs, but its effectiveness hinges on the efficiency and the reliability of the contextual CoT generation. Therefore, we propose Upfront CoT (UCoT), an efficient post-reasoning framework for CoT compression. UCoT trains a lightweight model (compressor) to provide contextual CoT in form of soft tokens and trains the LLM (executor) to leverage this contextual CoT for producing the final answer. Extensive experiments show that UCoT maintains the powerful reasoning ability of executor while significantly reducing the length of CoT. It is worth mentioning that when applying UCoT to the Qwen2.5-7B-Instruct model, the usage of tokens on GSM8K dataset is reduced by 50%, while the performance is 3.08% higher than that of the state-of-the-art (SOTA) method.

cs.CL↗
arXiv2025

MGT-Prism: Enhancing Domain Generalization for Machine-Generated Text Detection via Spectral Alignment

Large Language Models have shown growing ability to generate fluent and coherent texts that are highly similar to the writing style of humans. Current detectors for Machine-Generated Text (MGT) perform well when they are trained and tested in the same domain but generalize poorly to unseen domains, due to domain shift between data from different sources. In this work, we propose MGT-Prism, an MGT detection method from the perspective of the frequency domain for better domain generalization. Our key insight stems from analyzing text representations in the frequency domain, where we observe consistent spectral patterns across diverse domains, while significant discrepancies in magnitude emerge between MGT and human-written texts (HWTs). The observation initiates the design of a low frequency domain filtering module for filtering out the document-level features that are sensitive to domain shift, and a dynamic spectrum alignment strategy to extract the task-specific and domain-invariant features for improving the detector's performance in domain generalization. Extensive experiments demonstrate that MGT-Prism outperforms state-of-the-art baselines by an average of 0.90% in accuracy and 0.92% in F1 score on 11 test datasets across three domain-generalization scenarios.

cs.CL↗
arXiv2024

SL(n) covariant matrix-valued valuations on Orlicz spaces

All continuous, SL(n) covariant valuations on Orlicz spaces are completely classified without any symmetric assumptions. It is shown that the moment matrix is the only such valuation if n\geq3, while a new functional shows up in dimension two.

math.DG↗
arXiv2024

SL(n) covariant matrix-valued valuations on Lp-spaces

A complete classification is established for continuous and SL(n) covariant matrix-valued valuations on Lp(Rn,|x|2dx). The assumption of matrix symmetry is eliminated. For n>2, such valuation is uniquely characterized by the moment matrix of measurable function. In the 2-dimensional case, while the rotation matrix shows up.

math.DG↗
arXiv2024

Concentrate Attention: Towards Domain-Generalizable Prompt Optimization for Language Models

Recent advances in prompt optimization have notably enhanced the performance of pre-trained language models (PLMs) on downstream tasks. However, the potential of optimized prompts on domain generalization has been under-explored. To explore the nature of prompt generalization on unknown domains, we conduct pilot experiments and find that (i) Prompts gaining more attention weight from PLMs' deep layers are more generalizable and (ii) Prompts with more stable attention distributions in PLMs' deep layers are more generalizable. Thus, we offer a fresh objective towards domain-generalizable prompts optimization named "Concentration", which represents the "lookback" attention from the current decoding token to the prompt tokens, to increase the attention strength on prompts and reduce the fluctuation of attention distribution. We adapt this new objective to popular soft prompt and hard prompt optimization methods, respectively. Extensive experiments demonstrate that our idea improves comparison prompt optimization methods by 1.42% for soft prompt generalization and 2.16% for hard prompt generalization in accuracy on the multi-source domain generalization setting, while maintaining satisfying in-domain performance. The promising results validate the effectiveness of our proposed prompt optimization objective and provide key insights into domain-generalizable prompts.

cs.CL↗
arXiv2024

StablePT: Towards Stable Prompting for Few-shot Learning via Input Separation

Large language models have shown their ability to become effective few-shot learners with prompting, revolutionizing the paradigm of learning with data scarcity. However, this approach largely depends on the quality of prompt initialization, and always exhibits large variability among different runs. Such property makes prompt tuning highly unreliable and vulnerable to poorly constructed prompts, which limits its extension to more real-world applications. To tackle this issue, we propose to treat the hard prompt and soft prompt as separate inputs to mitigate noise brought by the prompt initialization. Furthermore, we optimize soft prompts with contrastive learning for utilizing class-aware information in the training process to maintain model performance. Experimental results demonstrate that \sysname outperforms state-of-the-art methods by 6.97% in accuracy and reduces the standard deviation by 1.92 on average. Furthermore, extensive experiments underscore its robustness and stability across 8 datasets covering various tasks. Codes are available at https://github.com/lccc0528/Stable/tree/main.

cs.CL↗
arXiv2024

Does DetectGPT Fully Utilize Perturbation? Bridging Selective Perturbation to Fine-tuned Contrastive Learning Detector would be Better

The burgeoning generative capabilities of large language models (LLMs) have raised growing concerns about abuse, demanding automatic machine-generated text detectors. DetectGPT, a zero-shot metric-based detector, first introduces perturbation and shows great performance improvement. However, in DetectGPT, the random perturbation strategy could introduce noise, and logit regression depends on the threshold, harming the generalizability and applicability of individual or small-batch inputs. Hence, we propose a novel fine-tuned detector, Pecola, bridging metric-based and fine-tuned methods by contrastive learning on selective perturbation. Selective strategy retains important tokens during perturbation and weights for multi-pair contrastive learning. The experiments show that Pecola outperforms the state-of-the-art (SOTA) by 1.20% in accuracy on average on four public datasets. And we further analyze the effectiveness, robustness, and generalization of the method.

cs.CL↗
arXiv2023

Doping dependence of superconductivity on a honeycomb lattice within the framework of kinetic-energy-driven superconductivity

Unconventional superconductivity on a honeycomb lattice has received increasing interest since the discovery of graphene primarily due to the similarities between materials with a honeycomb lattice and cuprate superconductors. Many theoretical studies have been conducted on superconductivity on a honeycomb lattice, however, a consistent picture is still lacking. In this article we have extended the theory of kinetic-energy-driven superconductivity, which has been developed to investigate unconventional superconductivity in cuprate superconductors, to explore superconductivity on a honeycomb lattice within the $t$-$J$ model. Our results demonstrate that the charge-carrier pair gap parameter with $d_{x^{2}-y^{2}}+{\rm i}d_{xy}$-wave symmetry exhibits a dome-like shape as a function of doping, with superconductivity emerging at a certain doping concentration and disappearing at high doping levels, similar to what has been observed in cuprate and cobaltate superconductors. Furthermore, the charge-carrier pair gap parameter decreases with increasing the value of $J/t$ (the antiferromagnetic exchange coupling constant relative to the nearest-neighbor hopping integral), and approaches zero when $J/t$ reaches a sufficiently large value. This indicates that the antiferromagnetic order will suppress the superconducting state and a sufficiently strong exchange coupling will completely destroy the superconductivity. Taking into account our present results together with the corresponding results of cuprate and cobaltate superconductors, it appears that the dome-like shape of the doping dependence of the charge-carrier pair gap parameter may be a common feature in doped Mott insulators.

cond-mat.supr-con↗
arXiv2023

Dialogue for Prompting: a Policy-Gradient-Based Discrete Prompt Generation for Few-shot Learning

Prompt-based pre-trained language models (PLMs) paradigm have succeeded substantially in few-shot natural language processing (NLP) tasks. However, prior discrete prompt optimization methods require expert knowledge to design the base prompt set and identify high-quality prompts, which is costly, inefficient, and subjective. Meanwhile, existing continuous prompt optimization methods improve the performance by learning the ideal prompts through the gradient information of PLMs, whose high computational cost, and low readability and generalizability are often concerning. To address the research gap, we propose a Dialogue-comprised Policy-gradient-based Discrete Prompt Optimization ($DP_2O$) method. We first design a multi-round dialogue alignment strategy for readability prompt set generation based on GPT-4. Furthermore, we propose an efficient prompt screening metric to identify high-quality prompts with linear complexity. Finally, we construct a reinforcement learning (RL) framework based on policy gradients to match the prompts to inputs optimally. By training a policy network with only 0.67% of the PLM parameter size on the tasks in the few-shot setting, $DP_2O$ outperforms the state-of-the-art (SOTA) method by 1.52% in accuracy on average on four open-source datasets. Moreover, subsequent experiments also demonstrate that $DP_2O$ has good universality, robustness, and generalization ability.

cs.LG↗
arXiv2023

Robust Approximate Dynamic Programming for Large-scale Unit Commitment with Energy Storages

The multistage robust unit commitment (UC) is of paramount importance for achieving reliable operations considering the uncertainty of renewable realizations. The typical affine decision rule method and the robust feasible region method may achieve uneconomic dispatches as the dispatch decisions just rely on the current-stage information. Through approximating the future cost-to-go functions, the dual dynamic programming based methods have been shown adaptive to the multistage robust optimization problems, while suffering from high computational complexity. Thus, we propose the robust approximate dynamic programming (RADP) method to promote the computational speed and the economic performance for large-scale robust UC problems. RADP initializes the candidate points for guaranteeing the feasibility of upper bounding the value functions, solves the linear McCormick relaxation based bilinear programming to obtain the worst cases, and combines the primal and dual updates for this hybrid binary and continuous decision-making problem to achieve fast convergence. We can verify that the RADP method enjoys a finite termination guarantee for the multistage robust optimization problems with achieving suboptimal solutions. Numerical tests on 118-bus and 2383-bus transmission systems have demonstrated that RADP can approach the suboptimal economic performance at significantly improved computational efficiency.

math.OC↗
arXiv2022

CoCo: Coherence-Enhanced Machine-Generated Text Detection Under Data Limitation With Contrastive Learning

Machine-Generated Text (MGT) detection, a task that discriminates MGT from Human-Written Text (HWT), plays a crucial role in preventing misuse of text generative models, which excel in mimicking human writing style recently. Latest proposed detectors usually take coarse text sequences as input and fine-tune pretrained models with standard cross-entropy loss. However, these methods fail to consider the linguistic structure of texts. Moreover, they lack the ability to handle the low-resource problem which could often happen in practice considering the enormous amount of textual data online. In this paper, we present a coherence-based contrastive learning model named CoCo to detect the possible MGT under low-resource scenario. To exploit the linguistic feature, we encode coherence information in form of graph into text representation. To tackle the challenges of low data resource, we employ a contrastive learning framework and propose an improved contrastive loss for preventing performance degradation brought by simple samples. The experiment results on two public datasets and two self-constructed datasets prove our approach outperforms the state-of-art methods significantly. Also, we surprisingly find that MGTs originated from up-to-date language models could be easier to detect than these from previous models, in our experiments. And we propose some preliminary explanations for this counter-intuitive phenomena. All the codes and datasets are open-sourced.

cs.CL↗
arXiv2020

Peak-structure in self-energy of cuprate superconductors

The recently deduced normal and anomalous self-energies from photoemission spectra of cuprate superconductors via the machine learning technique are calling for an explanation. Here the normal and anomalous self-energies in cuprate superconductors are analyzed within the framework of the kinetic-energy-driven superconductivity. It is shown that the exchanged spin excitations give rise to the well-pronounced low-energy peak-structures in both the normal and anomalous self-energies, however, they do not cancel in the total self-energy. In particular, the peak-structure in the normal self-energy is mainly responsible for the peak-dip-hump structure in the single-particle excitation spectrum, and can persist into the normal-state, while the sharp peak in the anomalous self-energy gives rise to a crucial contribution to the superconducting gap, and vanishes in the normal-state. Moreover, the evolution of the peak-structure with doping and momentum are also analyzed.

cond-mat.supr-con↗
arXiv2020

Renormalization of electrons in bilayer cuprate superconductors

The characteristic features of the renormalization of the electrons in the bilayer cuprate superconductors are investigated within the kinetic-energy driven superconductivity. It is shown that the quasiparticle excitation spectrum is split into its bonding and antibonding components due to the presence of the bilayer coupling, with each component that is independent. However, in the underdoped and optimally doped regimes, although the bonding and antibonding electron Fermi surface (EFS) contours deriving from the bonding and antibonding layers are truncated to form the bonding and antibonding Fermi arcs, almost all spectral weights in the bonding and antibonding Fermi arcs are reduced to the tips of the bonding and antibonding Fermi arcs, which in this case coincide with the bonding and antibonding hot spots. These hot spots connected by the scattering wave vectors ${\bf q}_{i} $ construct an octet scattering model, and then the enhancement of the quasiparticle scattering processes with the scattering wave vectors ${\bf q}_{i}$ is confirmed via the result of the autocorrelation of the ARPES spectral intensities. Moreover, the peak-dip-hump (PDH) structure developed in each component of the quasiparticle excitation spectrum along the corresponding EFS is directly related with the peak structure in the quasiparticle scattering rate except for at around the hot spots, where the PDH structure is caused mainly by the bilayer coupling. Although the kink in the quasiparticle dispersion is present all around EFS, when the momentum moves away from the node to the antinode, the kink energy smoothly decreases, while the dispersion kink becomes more pronounced, and in particular, near the cut close to the antinode, develops into a break separating of the fasting dispersing high-energy part of the quasiparticle excitation spectrum from the slower dispersing low-energy part.

cond-mat.supr-con↗