arXiv subjects

Ke Zeng

Publications and source records attributed to Ke Zeng.

51 records · Page 3Linked to original sources

arXiv2025

Reasoner for Real-World Event Detection: Scaling Reinforcement Learning via Adaptive Perplexity-Aware Sampling Strategy

Detecting abnormal events in real-world customer service dialogues is highly challenging due to the complexity of business data and the dynamic nature of customer interactions. Moreover, models must demonstrate strong out-of-domain (OOD) generalization to enable rapid adaptation across different business scenarios and maximize commercial value. In this work, we propose a novel Adaptive Perplexity-Aware Reinforcement Learning (APARL) framework that leverages the advanced reasoning capabilities of large language models for abnormal event detection. APARL introduces a dual-loop dynamic curriculum learning architecture, enabling the model to progressively focus on more challenging samples as its proficiency increases. This design effectively addresses performance bottlenecks and significantly enhances OOD transferability. Extensive evaluations on food delivery dialogue tasks show that our model achieves significantly enhanced adaptability and robustness, attaining the highest F1 score with an average improvement of 17.19\%, and an average improvement of 9.59\% in OOD transfer tests. This method provides a superior solution for industrial deployment of anomaly detection models, contributing to improved operational efficiency and commercial benefits.

cs.LG↗

arXiv2025

Don't Half-listen: Capturing Key-part Information in Continual Instruction Tuning

Instruction tuning for large language models (LLMs) can drive them to produce results consistent with human goals in specific downstream tasks. However, the process of continual instruction tuning (CIT) for LLMs may bring about the catastrophic forgetting (CF) problem, where previously learned abilities are degraded. Recent methods try to alleviate the CF problem by modifying models or replaying data, which may only remember the surface-level pattern of instructions and get confused on held-out tasks. In this paper, we propose a novel continual instruction tuning method based on Key-part Information Gain (KPIG). Our method computes the information gain on masked parts to dynamically replay data and refine the training objective, which enables LLMs to capture task-aware information relevant to the correct response and alleviate overfitting to general descriptions in instructions. In addition, we propose two metrics, P-score and V-score, to measure the generalization and instruction-following abilities of LLMs. Experiments demonstrate our method achieves superior performance on both seen and held-out tasks.

cs.CL↗

arXiv2025

When to Continue Thinking: Adaptive Thinking Mode Switching for Efficient Reasoning

Large reasoning models (LRMs) achieve remarkable performance via long reasoning chains, but often incur excessive computational overhead due to redundant reasoning, especially on simple tasks. In this work, we systematically quantify the upper bounds of LRMs under both Long-Thinking and No-Thinking modes, and uncover the phenomenon of "Internal Self-Recovery Mechanism" where models implicitly supplement reasoning during answer generation. Building on this insight, we propose Adaptive Self-Recovery Reasoning (ASRR), a framework that suppresses unnecessary reasoning and enables implicit recovery. By introducing accuracy-aware length reward regulation, ASRR adaptively allocates reasoning effort according to problem difficulty, achieving high efficiency with negligible performance sacrifice. Experiments across multiple benchmarks and models show that, compared with GRPO, ASRR reduces reasoning budget by up to 32.5% (1.5B) and 25.7% (7B) with minimal accuracy loss (1.2% and 0.6% pass@1), and significantly boosts harmless rates on safety benchmarks (up to +21.7%). Our results highlight the potential of ASRR for enabling efficient, adaptive, and safer reasoning in LRMs.

cs.AI↗

arXiv2025

Restoring Heisenberg-Limited Precision in Non-Markovian Open Quantum Systems via Dynamical Decoupling

Non-classical resources enable measurements to achieve a precision that exceeds the limits predicted by the central limit theorem. However, environmental noise arising from system-environment interactions severely limits the performance of such resources through decoherence. While significant progress has been made in mitigating Markovian noise, the extent to which non-Markovian noise can be mitigated remains poorly understood. We demonstrate that Heisenberg Scaling, the ultimate quantum limit on measurement precision, can be recovered in quantum metrology under non-Markovian noise by leveraging carefully designed Dynamical Decoupling Techniques. Importantly, our approach does not rely on assumptions of Markovian dynamics. By imposing appropriate conditions on the control Hamiltonian, we show that HS can be achieved irrespective of whether the noise is Markovian or non-Markovian. We also prove necessary and sufficient conditions for the existence of such control Hamiltonians. As an illustrative example, we apply our framework to the damped Jaynes-Cummings model, successfully mitigating memory effects and maintaining measurement precision in complex, non-Markovian environments. These findings highlight the power of quantum control to overcome decoherence challenges and enhance metrological performance in realistic, noisy quantum systems.

quant-ph↗

arXiv2024

Learning or Self-aligning? Rethinking Instruction Fine-tuning

Instruction Fine-tuning~(IFT) is a critical phase in building large language models~(LLMs). Previous works mainly focus on the IFT's role in the transfer of behavioral norms and the learning of additional world knowledge. However, the understanding of the underlying mechanisms of IFT remains significantly limited. In this paper, we design a knowledge intervention framework to decouple the potential underlying factors of IFT, thereby enabling individual analysis of different factors. Surprisingly, our experiments reveal that attempting to learn additional world knowledge through IFT often struggles to yield positive impacts and can even lead to markedly negative effects. Further, we discover that maintaining internal knowledge consistency before and after IFT is a critical factor for achieving successful IFT. Our findings reveal the underlying mechanisms of IFT and provide robust support for some very recent and potential future works.

cs.CL↗

arXiv2024

CLAQ: Pushing the Limits of Low-Bit Post-Training Quantization for LLMs

Parameter quantization for Large Language Models (LLMs) has attracted increasing attentions recently in reducing memory costs and improving computational efficiency. Early approaches have been widely adopted. However, the existing methods suffer from poor performance in low-bit (such as 2 to 3 bits) scenarios. In this paper, we present a novel and effective Column-Level Adaptive weight Quantization (CLAQ) framework by introducing three different types of adaptive strategies for LLM quantization. Firstly, a K-Means clustering based algorithm is proposed that allows dynamic generation of quantization centroids for each column of a parameter matrix. Secondly, we design an outlier-guided adaptive precision search strategy which can dynamically assign varying bit-widths to different columns. Finally, a dynamic outlier reservation scheme is developed to retain some parameters in their original float point precision, in trade off of boosted model performance. Experiments on various mainstream open source LLMs including LLaMA-1, LLaMA-2 and Yi demonstrate that our methods achieve the state-of-the-art results across different bit settings, especially in extremely low-bit scenarios. Code is available at https://github.com/fayuge/CLAQ.

cs.LG↗

arXiv2024

One-Shot Sensitivity-Aware Mixed Sparsity Pruning for Large Language Models

Various Large Language Models~(LLMs) from the Generative Pretrained Transformer(GPT) family have achieved outstanding performances in a wide range of text generation tasks. However, the enormous model sizes have hindered their practical use in real-world applications due to high inference latency. Therefore, improving the efficiencies of LLMs through quantization, pruning, and other means has been a key issue in LLM studies. In this work, we propose a method based on Hessian sensitivity-aware mixed sparsity pruning to prune LLMs to at least 50% sparsity without the need of any retraining. It allocates sparsity adaptively based on sensitivity, allowing us to reduce pruning-induced error while maintaining the overall sparsity level. The advantages of the proposed method exhibit even more when the sparsity is extremely high. Furthermore, our method is compatible with quantization, enabling further compression of LLMs. We have released the available code.

cs.CL↗

arXiv2023

A Task-oriented Dialog Model with Task-progressive and Policy-aware Pre-training

Pre-trained conversation models (PCMs) have achieved promising progress in recent years. However, existing PCMs for Task-oriented dialog (TOD) are insufficient for capturing the sequential nature of the TOD-related tasks, as well as for learning dialog policy information. To alleviate these problems, this paper proposes a task-progressive PCM with two policy-aware pre-training tasks. The model is pre-trained through three stages where TOD-related tasks are progressively employed according to the task logic of the TOD system. A global policy consistency task is designed to capture the multi-turn dialog policy sequential relation, and an act-based contrastive learning task is designed to capture similarities among samples with the same dialog policy. Our model achieves better results on both MultiWOZ and In-Car end-to-end dialog modeling benchmarks with only 18\% parameters and 25\% pre-training data compared to the previous state-of-the-art PCM, GALAXY.

cs.CL↗

arXiv2023

Wide-band Unambiguous Quantum Sensing via Geodesic Evolution

We present a quantum sensing technique that utilizes a sequence of $π$ pulses to cyclically drive the qubit dynamics along a geodesic path of adiabatic evolution. This approach effectively suppresses the effects of both decoherence noise and control errors while simultaneously removing unwanted resonance terms, such as higher harmonics and spurious responses commonly encountered in dynamical decoupling control. As a result, our technique offers robust, wide-band, unambiguous, and high-resolution quantum sensing capabilities for signal detection and individual addressing of quantum systems, including spins. To demonstrate its versatility, we showcase successful applications of our method in both low-frequency and high-frequency sensing scenarios. The significance of this quantum sensing technique extends to the detection of complex signals and the control of intricate quantum environments. By enhancing detection accuracy and enabling precise manipulation of quantum systems, our method holds considerable promise for a variety of practical applications.

quant-ph↗

arXiv2018

Generative discriminative models for multivariate inference and statistical mapping in medical imaging

This paper presents a general framework for obtaining interpretable multivariate discriminative models that allow efficient statistical inference for neuroimage analysis. The framework, termed generative discriminative machine (GDM), augments discriminative models with a generative regularization term. We demonstrate that the proposed formulation can be optimized in closed form and in dual space, allowing efficient computation for high dimensional neuroimaging datasets. Furthermore, we provide an analytic estimation of the null distribution of the model parameters, which enables efficient statistical inference and p-value computation without the need for permutation testing. We compared the proposed method with both purely generative and discriminative learning methods in two large structural magnetic resonance imaging (sMRI) datasets of Alzheimer's disease (AD) (n=415) and Schizophrenia (n=853). Using the AD dataset, we demonstrated the ability of GDM to robustly handle confounding variations. Using Schizophrenia dataset, we demonstrated the ability of GDM to handle multi-site studies. Taken together, the results underline the potential of the proposed approach for neuroimaging analyses.

stat.AP↗

arXiv2017

Entropic uncertainty relation under quantum channels with memory

Recently, Xu et al. [Phys. Rev. A 86, 012113(2012)] explored the behavior of the entropic uncertainty relation under the influence of local unital and nonunital noisy channels for a class of Bell-diagonal states. We here reform their results and investigate the entropic uncertainty relation under the influence of unital and nonunital noisy channels with memory. Different types of noisy channels with memory, such as amplitude damping channel(nonunitary), phase-damping and depolarizing channels(unitary) have been taken into account. Some analytical or numerical results are presented. The effect of channels with memory on dynamics of the entropic uncertainties (or their lower bounds) has been discussed in detail. Compare with previous results, our results show that, the entropic uncertainties (or their lower bounds) subjecting to amplitude damping channel with memory will be reduced at first and then be lifted with the memory coefficient of channel $μ$ increasing, however they will be only reduced under phase-damping and depolarizing channels with memory. Especially, in the limit of $μ\rightarrow1$, the entropic uncertainties (or their lower bounds) could be well protected and immune to decoherence of channle. Moreover, the mechanism behind these phenomena are also explored by using the purity of state.

quant-ph↗

arXiv2016

Quantum coherence of two-qubit over quantum channels with memory

Using the axiomatic definition of the coherence measure, such as the $l_{1}$ norm and the relative entropy, we study the phenomena of two-qubit system quantum coherence through quantum channels where successive uses of the channels are memory. Different types of noisy channels with memory, such as amplitude damping, phase-damping, and depolarizing channels effect on quantum coherence have been discussed in detail. The results show that, quantum channels with memory can efficiently protect coherence from noisy channels. Particularly, as channels with perfect memory, quantum coherence is unaffected by the phase damping as well as depolarizing channels. Besides, we also investigate the cohering and decohering power of quantum channels with memory.

quant-ph↗

arXiv2016

Non-Hermitian-assisted enhancement precision of parameter estimation

Recently, Zhong et al. [Phys. Rev. A 87, 022337(2013)] investigated the dynamics of quantum Fisher information (QFI) in the presence of decoherence. We here reform their results and propose two schemes to enhance and preserve the QFIs for a qubit system subjected to a decoherence noisy environment by applying $\textit{non-Hermitian}$ operators process. The proposed schemes are related to the $\textit{prior}$ (or the $\textit{post}$ ) non-Hermitian operator process corresponding to before (or after) the interest system suffers from decoherence, respectively. Resorting to the Bloch sphere representation, we derive the exact analytical expressions of the QFIs with respect to the amplitude parameter $θ$ and the phase parameter $ϕ$, and in detail investigate the influence of $\textit{non-Hermitian}$ operator parameters on the QFIs. Compared with purity decoherence process (without non-Hermitian operator process), we find that the $\textit{post non-Hermitian}$ operator process can potentially enhance and preserve the QFIs by choosing appropriate $\textit{non-Hermitian}$ operator parameters, while with the help of the $\textit{prior non-Hermitian}$ operator process one could completely eliminate the effect of decoherence to improve the parameters estimation. Finally, a generalized non-Hermitian operator parameters effect on the parameters estimation is also considered.

quant-ph↗

arXiv2016

Generation and protection of steady-state quantum correlations due to quantum channels with memory

We have proposed a scheme of the generation and preservation of two-qubit steady state quantum correlations through quantum channels where successive uses of the channels are correlated. Different types of noisy channels with memory, such as amplitude damping, phase-damping, and depolarizing channels have been taken into account. Some analytical or numerical results are presented. The effect of channels with memory on dynamics of quantum correlations has been discussed in detail. The results show that, steady state entanglement between two independent qubits without entanglement subject to amplitude damping channel with memory can be generated. Besides, we compare the dynamics of entanglement with that of quantum discord when a two-qubit system is prepared in an entangled state. We show that entanglement dynamics suddenly disappears, while quantum discord displays only in the asymptotic limit. Two-qubit quantum correlations can be preserved at a long time in the limit of $μ\rightarrow1$.

quant-ph↗

arXiv2015

Distillability sudden death and sudden birth in a two-qutrit system under decoherence of finite temperature

Distillability sudden death and sudden birth in a two-qutrit system under decoherence of finite temperature is studied in detail. By using of the negativity and realignment criterion, it is shown that certain initial prepared free entangled states may become bound entangled or separable states in a finite time. Moreover, initial prepared bound entangled or separable states also may become distillabile entangled states in a finite time.

quant-ph↗