arXiv subjects

Shimin Li

Publications and source records attributed to Shimin Li.

36 records · Page 2Linked to original sources

arXiv2024

Unified Active Retrieval for Retrieval Augmented Generation

In Retrieval-Augmented Generation (RAG), retrieval is not always helpful and applying it to every instruction is sub-optimal. Therefore, determining whether to retrieve is crucial for RAG, which is usually referred to as Active Retrieval. However, existing active retrieval methods face two challenges: 1. They usually rely on a single criterion, which struggles with handling various types of instructions. 2. They depend on specialized and highly differentiated procedures, and thus combining them makes the RAG system more complicated and leads to higher response latency. To address these challenges, we propose Unified Active Retrieval (UAR). UAR contains four orthogonal criteria and casts them into plug-and-play classification tasks, which achieves multifaceted retrieval timing judgements with negligible extra inference cost. We further introduce the Unified Active Retrieval Criteria (UAR-Criteria), designed to process diverse active retrieval scenarios through a standardized procedure. Experiments on four representative types of user instructions show that UAR significantly outperforms existing work on the retrieval timing judgement and the performance of downstream tasks, which shows the effectiveness of UAR and its helpfulness to downstream tasks.

cs.CL↗

arXiv2024

On the stability of singular Hopf bifurcation and its application

Recently, research on the complex periodic behavior of multi-scale systems has become increasingly popular. Krupa et al. \cite{krupa2} provided a way to obtain relaxation oscillations in slow-fast systems through singular Hopf bifurcations and canard explosion. The authors derived a $O(1)$ expression $A$ for the first Lyapunov coefficient (under the condition $A \neq 0$), and deduced the bifurcation curves of singular Hopf and canard explosions. This paper employs Blow-up technique, normal form theory, and Lyapunov coefficient formula to present higher-order approximate expressions for the first Lyapunov coefficient when $A=0$ for slow-fast systems. As an application, we investigate the bifurcation phenomena of a predator-prey model with Allee effects. Utilizing the formulas obtained in this paper, we identify both supercritical and subcritical Hopf bifurcations that may occur simultaneously in the system. Numerical simulations validate the results. Finally, by normal form and slow divergence integral theory, we prove the cyclicity of the system is 1.

math.DS↗

arXiv2024

SpeechAlign: Aligning Speech Generation to Human Preferences

Speech language models have significantly advanced in generating realistic speech, with neural codec language models standing out. However, the integration of human feedback to align speech outputs to human preferences is often neglected. This paper addresses this gap by first analyzing the distribution gap in codec language models, highlighting how it leads to discrepancies between the training and inference phases, which negatively affects performance. Then we explore leveraging learning from human feedback to bridge the distribution gap. We introduce SpeechAlign, an iterative self-improvement strategy that aligns speech language models to human preferences. SpeechAlign involves constructing a preference codec dataset contrasting golden codec tokens against synthetic tokens, followed by preference optimization to improve the codec language model. This cycle of improvement is carried out iteratively to steadily convert weak models to strong ones. Through both subjective and objective evaluations, we show that SpeechAlign can bridge the distribution gap and facilitating continuous self-improvement of the speech language model. Moreover, SpeechAlign exhibits robust generalization capabilities and works for smaller models. Code and models will be available at https://github.com/0nutation/SpeechGPT.

cs.CL↗

arXiv2024

Agent Alignment in Evolving Social Norms

Agents based on Large Language Models (LLMs) are increasingly permeating various domains of human production and life, highlighting the importance of aligning them with human values. The current alignment of AI systems primarily focuses on passively aligning LLMs through human intervention. However, agents possess characteristics like receiving environmental feedback and self-evolution, rendering the LLM alignment methods inadequate. In response, we propose an evolutionary framework for agent evolution and alignment, named EvolutionaryAgent, which transforms agent alignment into a process of evolution and selection under the principle of survival of the fittest. In an environment where social norms continuously evolve, agents better adapted to the current social norms will have a higher probability of survival and proliferation, while those inadequately aligned dwindle over time. Experimental results assessing the agents from multiple perspectives in aligning with social norms demonstrate that EvolutionaryAgent can align progressively better with the evolving social norms while maintaining its proficiency in general tasks. Effectiveness tests conducted on various open and closed-source LLMs as the foundation for agents also prove the applicability of our approach.

cs.CL↗

arXiv2024

LLM can Achieve Self-Regulation via Hyperparameter Aware Generation

In the realm of Large Language Models (LLMs), users commonly employ diverse decoding strategies and adjust hyperparameters to control the generated text. However, a critical question emerges: Are LLMs conscious of the existence of these decoding strategies and capable of regulating themselves? The current decoding generation process often relies on empirical and heuristic manual adjustments to hyperparameters based on types of tasks and demands. However, this process is typically cumbersome, and the decoding hyperparameters may not always be optimal for each sample. To address the aforementioned challenges, we propose a novel text generation paradigm termed Hyperparameter Aware Generation (HAG). By leveraging hyperparameter-aware instruction tuning, the LLM autonomously determines the optimal decoding strategy and configs based on the input samples, enabling self-regulation. Our approach eliminates the need for extensive manual tuning, offering a more autonomous, self-regulate model behavior. Experimental results spanning six datasets across reasoning, creativity, translation, and mathematics tasks demonstrate that hyperparameter-aware instruction tuning empowers the LLMs to self-regulate the decoding strategy and hyperparameter. HAG extends the current paradigm in the text generation process, highlighting the feasibility of endowing the LLMs with self-regulate decoding strategies.

cs.CL↗

arXiv2024

Can AI Assistants Know What They Don't Know?

Recently, AI assistants based on large language models (LLMs) show surprising performance in many tasks, such as dialogue, solving math problems, writing code, and using tools. Although LLMs possess intensive world knowledge, they still make factual errors when facing some knowledge intensive tasks, like open-domain question answering. These untruthful responses from the AI assistant may cause significant risks in practical applications. We believe that an AI assistant's refusal to answer questions it does not know is a crucial method for reducing hallucinations and making the assistant truthful. Therefore, in this paper, we ask the question "Can AI assistants know what they don't know and express them through natural language?" To answer this question, we construct a model-specific "I don't know" (Idk) dataset for an assistant, which contains its known and unknown questions, based on existing open-domain question answering datasets. Then we align the assistant with its corresponding Idk dataset and observe whether it can refuse to answer its unknown questions after alignment. Experimental results show that after alignment with Idk datasets, the assistant can refuse to answer most its unknown questions. For questions they attempt to answer, the accuracy is significantly higher than before the alignment.

cs.CL↗

arXiv2024

SpeechGPT-Gen: Scaling Chain-of-Information Speech Generation

Benefiting from effective speech modeling, current Speech Large Language Models (SLLMs) have demonstrated exceptional capabilities in in-context speech generation and efficient generalization to unseen speakers. However, the prevailing information modeling process is encumbered by certain redundancies, leading to inefficiencies in speech generation. We propose Chain-of-Information Generation (CoIG), a method for decoupling semantic and perceptual information in large-scale speech generation. Building on this, we develop SpeechGPT-Gen, an 8-billion-parameter SLLM efficient in semantic and perceptual information modeling. It comprises an autoregressive model based on LLM for semantic information modeling and a non-autoregressive model employing flow matching for perceptual information modeling. Additionally, we introduce the novel approach of infusing semantic information into the prior distribution to enhance the efficiency of flow matching. Extensive experimental results demonstrate that SpeechGPT-Gen markedly excels in zero-shot text-to-speech, zero-shot voice conversion, and speech-to-speech dialogue, underscoring CoIG's remarkable proficiency in capturing and modeling speech's semantic and perceptual dimensions. Code and models are available at https://github.com/0nutation/SpeechGPT.

cs.CL↗

arXiv2024

SpeechTokenizer: Unified Speech Tokenizer for Speech Large Language Models

Current speech large language models build upon discrete speech representations, which can be categorized into semantic tokens and acoustic tokens. However, existing speech tokens are not specifically designed for speech language modeling. To assess the suitability of speech tokens for building speech language models, we established the first benchmark, SLMTokBench. Our results indicate that neither semantic nor acoustic tokens are ideal for this purpose. Therefore, we propose SpeechTokenizer, a unified speech tokenizer for speech large language models. SpeechTokenizer adopts the Encoder-Decoder architecture with residual vector quantization (RVQ). Unifying semantic and acoustic tokens, SpeechTokenizer disentangles different aspects of speech information hierarchically across different RVQ layers. Furthermore, We construct a Unified Speech Language Model (USLM) leveraging SpeechTokenizer. Experiments show that SpeechTokenizer performs comparably to EnCodec in speech reconstruction and demonstrates strong performance on the SLMTokBench benchmark. Also, USLM outperforms VALL-E in zero-shot Text-to-Speech tasks. Code and models are available at https://github.com/ZhangXInFD/SpeechTokenizer/.

cs.CL↗

arXiv2023

On Choosing Initial Values of Iteratively Reweighted $\ell_1$ Algorithms for the Piece-wise Exponential Penalty

Computing the proximal operator of the sparsity-promoting piece-wise exponential (PiE) penalty $1-e^{-|x|/σ}$ with a given shape parameter $σ>0$, which is treated as a popular nonconvex surrogate of $\ell_0$-norm, is fundamental in feature selection via support vector machines, image reconstruction, zero-one programming problems, compressed sensing, etc. Due to the nonconvexity of PiE, for a long time, its proximal operator is frequently evaluated via an iteratively reweighted $\ell_1$ algorithm, which substitutes PiE with its first-order approximation, however, the obtained solutions only are the critical point. Based on the exact characterization of the proximal operator of PiE, we explore how the iteratively reweighted $\ell_1$ solution deviates from the true proximal operator in certain regions, which can be explicitly identified in terms of $σ$, the initial value and the regularization parameter in the definition of the proximal operator. Moreover, the initial value can be adaptively and simply chosen to ensure that the iteratively reweighted $\ell_1$ solution belongs to the proximal operator of PiE.

math.NA↗

arXiv2023

Multijugate Dual Learning for Low-Resource Task-Oriented Dialogue System

Dialogue data in real scenarios tend to be sparsely available, rendering data-starved end-to-end dialogue systems trained inadequately. We discover that data utilization efficiency in low-resource scenarios can be enhanced by mining alignment information uncertain utterance and deterministic dialogue state. Therefore, we innovatively implement dual learning in task-oriented dialogues to exploit the correlation of heterogeneous data. In addition, the one-to-one duality is converted into a multijugate duality to reduce the influence of spurious correlations in dual training for generalization. Without introducing additional parameters, our method could be implemented in arbitrary networks. Extensive empirical analyses demonstrate that our proposed method improves the effectiveness of end-to-end task-oriented dialogue systems under multiple benchmarks and obtains state-of-the-art results in low-resource scenarios.

cs.CL↗

arXiv2023

SpeechGPT: Empowering Large Language Models with Intrinsic Cross-Modal Conversational Abilities

Multi-modal large language models are regarded as a crucial step towards Artificial General Intelligence (AGI) and have garnered significant interest with the emergence of ChatGPT. However, current speech-language models typically adopt the cascade paradigm, preventing inter-modal knowledge transfer. In this paper, we propose SpeechGPT, a large language model with intrinsic cross-modal conversational abilities, capable of perceiving and generating multi-model content. With discrete speech representations, we first construct SpeechInstruct, a large-scale cross-modal speech instruction dataset. Additionally, we employ a three-stage training strategy that includes modality-adaptation pre-training, cross-modal instruction fine-tuning, and chain-of-modality instruction fine-tuning. The experimental results demonstrate that SpeechGPT has an impressive capacity to follow multi-modal human instructions and highlight the potential of handling multiple modalities with one model. Demos are shown in https://0nutation.github.io/SpeechGPT.github.io/.

cs.CL↗

arXiv2022

Mitigating Negative Style Transfer in Hybrid Dialogue System

As the functionality of dialogue systems evolves, hybrid dialogue systems that accomplish user-specific goals and participate in open-topic chitchat with users are attracting growing attention. Existing research learns both tasks concurrently utilizing a multi-task fusion technique but ignores the negative transfer phenomenon induced by the unique textual style differences. Therefore, contrastive learning based on the latent variable model is used to decouple the various textual genres in the latent space. We devise supervised and self-supervised positive and negative sample constructions for diverse datasets. In addition, to capitalize on the style information contained in the decoupled latent variables, we employ a style prefix that incorporates latent variables further to control the generation of responses with varying styles. We performed extensive experiments on three dialogue datasets, including a hybrid dialogue dataset and two task-oriented dialogue datasets. The experimental results demonstrate that our method can mitigate the negative style transfer issue and achieves state-of-the-art performance on multiple dialogue datasets.

cs.CL↗

arXiv2022

Contrast and Generation Make BART a Good Dialogue Emotion Recognizer

In dialogue systems, utterances with similar semantics may have distinctive emotions under different contexts. Therefore, modeling long-range contextual emotional relationships with speaker dependency plays a crucial part in dialogue emotion recognition. Meanwhile, distinguishing the different emotion categories is non-trivial since they usually have semantically similar sentiments. To this end, we adopt supervised contrastive learning to make different emotions mutually exclusive to identify similar emotions better. Meanwhile, we utilize an auxiliary response generation task to enhance the model's ability of handling context information, thereby forcing the model to recognize emotions with similar semantics in diverse contexts. To achieve these objectives, we use the pre-trained encoder-decoder model BART as our backbone model since it is very suitable for both understanding and generation tasks. The experiments on four datasets demonstrate that our proposed model obtains significantly more favorable results than the state-of-the-art model in dialogue emotion recognition. The ablation study further demonstrates the effectiveness of supervised contrastive loss and generative loss.

cs.CL↗

arXiv2018

Algorithms for Covering Multiple Barriers

In this paper, we consider the problems for covering multiple intervals on a line. Given a set $B$ of $m$ line segments (called "barriers") on a horizontal line $L$ and another set $S$ of $n$ horizontal line segments of the same length in the plane, we want to move all segments of $S$ to $L$ so that their union covers all barriers and the maximum movement of all segments of $S$ is minimized. Previously, an $O(n^3\log n)$-time algorithm was given for the case $m=1$. In this paper, we propose an $O(n^2\log n\log \log n+nm\log m)$-time algorithm for a more general setting with any $m\geq 1$, which also improves the previous work when $m=1$. We then consider a line-constrained version of the problem in which the segments of $S$ are all initially on the line $L$. Previously, an $O(n\log n)$-time algorithm was known for the case $m=1$. We present an algorithm of $O(m\log m+n\log m \log n)$ time for any $m\geq 1$. These problems may have applications in mobile sensor barrier coverage in wireless sensor networks.

cs.CG↗

arXiv2017

Separating Overlapped Intervals on a Line

Given $n$ intervals on a line $\ell$, we consider the problem of moving these intervals on $\ell$ such that no two intervals overlap and the maximum moving distance of the intervals is minimized. The difficulty for solving the problem lies in determining the order of the intervals in an optimal solution. By interesting observations, we show that it is sufficient to consider at most $n$ "candidate" lists of ordered intervals. Further, although explicitly maintaining these lists takes $Ω(n^2)$ time and space, by more observations and a pruning technique, we present an algorithm that can compute an optimal solution in $O(n\log n)$ time and $O(n)$ space. We also prove an $Ω(n\log n)$ time lower bound for solving the problem, which implies the optimality of our algorithm.

cs.CG↗

arXiv2016

Dispersing Points on Intervals

We consider a problem of dispersing points on disjoint intervals on a line. Given n pairwise disjoint intervals sorted on a line, we want to find a point in each interval such that the minimum pairwise distance of these points is maximized. Based on a greedy strategy, we present a linear time algorithm for the problem. Further, we also solve in linear time the cycle version of the problem where the intervals are given on a cycle.

cs.CG↗

arXiv2016

On the number of limit cycles for a class of discontinuous quadratic differential systems

The present paper is devoted to the study of the maximum number of limit cycles bifurcated from the periodic orbits of the quadratic isochronous center $\dot{x}=-y+\frac{16}{3}x^{2}-\frac{4}{3}y^{2},\dot{y}=x+\frac{8}{3}xy$ by the averaging method of first order, when it is perturbed inside a class of discontinuous quadratic polynomial differential systems. The \emph{Chebyshev criterion} is used to show that this maximum number is 5 and can be realizable. The result and that in paper \cite{LC} completely answer the questions left in the paper \cite{LM}.

math.CA↗

arXiv2016

Center-focus determination and limit cycles bifurcation for $p:q$ homogeneous weight singular point

The quasi-homogeneous (and in general non-homogeneous) polynomial differential systems have been studied from many different points of view. In this paper, Center-focus determination and limit cycles bifurcation for $p:q$ homogeneous weight singular point are investigated. Some prosperities of Successive function and focus values are discussed, furthermore, the method of computing focal values is given. As an example, center-focus determination and limit cycle bifurcation for $2:3$ homogeneous weight singular point are studied, three or five limit cycles in the neighborhood of origin can be obtained by different perturbations.

math.DS↗