arXiv subjects

Yuyang Huang

Publications and source records attributed to Yuyang Huang.

32 records · Page 2Linked to original sources

arXiv2025

Fast and robust consensus-based optimization via optimal feedback control

We propose a variant of consensus-based optimization (CBO) algorithms, controlled-CBO, which introduces a feedback control term to improve convergence towards global minimizers of non-convex functions in multiple dimensions. The feedback law is a gradient of a numerical approximation to the Hamilton-Jacobi-Bellman (HJB) equation, which serves as a proxy of the original objective function. Thus, the associated control signal furnishes gradient-like information to facilitate the identification of the global minimum without requiring derivative computation from the objective function itself. The proposed method exhibits significantly improved performance over standard CBO methods in numerical experiments, particularly in scenarios involving a limited number of particles, or where the initial particle ensemble is not well positioned with respect to the global minimum. At the same time, the modification keeps the algorithm amenable to theoretical analysis in the mean-field sense. The superior convergence rates are assessed experimentally.

math.OC↗

arXiv2025

DroidSpeak: KV Cache Sharing for Cross-LLM Communication and Multi-LLM Serving

Compound AI systems, such as agentic systems, are an emerging trend in large-scale enterprise settings, with multiple LLMs specialized for different users, tasks, and/or roles working together. In these scenarios, different models often process inputs that share the same context prefix. Although much work was done in the past to enable the reuse of prefix KV caches across inputs for a single model, how to enable one model to reuse the prefix KV caches of a different model remains an open question. We introduce DroidSpeak, the first distributed LLM inference system that enables KV cache reuse across distributed nodes running inference of different LLMs, so long as the LLMs have the same architecture. We present the first study that aims at understanding the impact of sharing KV caches across different LLMs, and if/when such sharing affects quality. Inspired by the findings, we present DroidSpeak, which selectively recomputes a few layers of the KV cache produced by another LLM and reuses the remaining layers, with negligible quality loss. Moreover, carefully pipelining the layer-wise re-computation and the loading of reused KV cache further improves the inference performance. Experiments on diverse datasets and model pairs demonstrate that DroidSpeak achieves up to 4x throughput improvement and about 3.1x faster prefill (time to first token), with negligible loss of quality in F1 scores, Rouge-L or code similarity score, compared to the baseline which does not allow any sharing across models.

cs.MA↗

arXiv2025

A multiscale Consensus-Based algorithm for multi-level optimization

A novel multiscale consensus-based optimization (CBO) algorithm for solving bi- and tri-level optimization problems is introduced. Existing CBO techniques are generalized by the proposed method through the employment of multiple interacting populations of particles, each of which is used to optimize one level of the problem. These particle populations are evolved through multiscale-in-time dynamics, which are formulated as a singularly perturbed system of stochastic differential equations. Theoretical convergence analysis for the multiscale CBO model to an averaged effective dynamics as the time-scale separation parameter approaches zero is provided. The resulting algorithm is presented for both bi-level and tri-level optimization problems. The effectiveness of the approach in tackling complex multi-level optimization tasks is demonstrated through numerical experiments on various benchmark functions. Additionally, it is shown that the proposed method performs well on min-max optimization problems, comparing favorably with existing CBO algorithms for saddle point problems.

math.OC↗

arXiv2025

Procedure of tuning up a three-site artificial Kitaev chain based on transmon measurements

Artificial Kitaev chains (AKCs), formed of quantum dot-superconductor linear arrays, provide a promising platform for hosting Majorana bound states (MBSs) and implementing topological quantum computing. The main challenges along this research direction would include the tuning up of AKCs for hosting MBSs and the readout of the parity of the chains. In this work, we present a step-by-step procedure for tuning up a three-site AKC to its sweet spots based on the spectra of a transmon circuit which is integrated with the chain for the purpose of reading out the parity of the chain. The signatures of the transmon's plasma modes in each step, particular those related to the appearance of MBSs in the chain, will be given. We find that the sweet spots in a three-site AKC can be classified into three types based on the relative strengths of elastic cotunneling (ECT) and crossed Andreev reflection (CAR): ECT-dominated sweet spots, genuine sweet spots and CAR-dominated sweet spots. We show that the ECT-dominated and CAR-dominated sweet spots can be more conveniently accessed and utilized in transmon-based measurements.

cond-mat.mes-hall↗

arXiv2025

Measurement of parity-dependent energy-phase relation of the low-energy states in a potential artificial Kitaev chain utilizing a transmon qubit

Artificial Kitaev chains have emerged as a promising platform for realizing topological quantum computing. Once the chains are formed and the Majorana zero modes are braided/fused, reading out the parity of the chains is essential for further verifying the non-Abelian property of the Majorana zero modes. Here we demonstrate the feasibility of using a superconducting transmon qubit, which incorporates an end of a four-site quantum dot-superconductor chain based on a Ge/Si nanowire, to directly detect the singlet/doublet state, and thus the parity of the entire chain. We also demonstrate that for multiple-dot chains there are two types of 0-π transitions between different charging states: the parity-flip 0-π transition and the parity-preserved 0-π transition. Furthermore, we show that the inter-dot coupling, hence the strengths of cross Andreev reflection and elastic cotunneling of electrons, can be adjusted by local electrostatic gating in chains fabricated on Ge/Si core-shell nanowires. Our exploration would be helpful for the ultimate realization of topological quantum computing based on artificial Kitaev chains.

cond-mat.mes-hall↗

arXiv2025

Controllable creation of topological boundary states in topological-insulator-based Josephson corner junctions

Majorana zero modes (MZMs) in condensed matter systems have attracted great attention in the past two decades, due to their interesting physics and potential application in topological quantum computing (TQC). However, the topologically protected nature of MZMs still need more experimental verifications. In this study, we have realized controllable creation of a topological boundary state at the corner of topological insulator (TI)-based Josephson corner junctions. This state demonstrates protected existence across a broad region in parametric space, and exhibits a non-2π-period but 4π-period-compatible energy-phase relation. Our study suggests that TI-based Josephson junctions, as proposed in the Fu-Kane scheme of TQC, may provide a promising platform for hosting and braiding MZMs.

cond-mat.mes-hall↗

arXiv2025

Alchemist: Towards the Design of Efficient Online Continual Learning System

Continual learning has become a promising solution to refine large language models incrementally by leveraging user feedback. In particular, online continual learning - iteratively training the model with small batches of user feedback - has demonstrated notable performance improvements. However, the existing practice of separating training and serving processes forces the online trainer to recompute the intermediate results already done during serving. Such redundant computations can account for 30%-42% of total training time. In this paper, we propose Alchemist, to the best of our knowledge, the first online continual learning system that efficiently reuses serving activations to increase training throughput. Alchemist introduces two key techniques: (1) recording and storing activations and KV cache only during the prefill phase to minimize latency and memory overhead; and (2) smart activation offloading and hedging. Evaluations with inputs of varied token length sampled from ShareGPT dataset show that compared with a separate training cluster, Alchemist significantly increases training throughput by up to 1.72x, reduces up to 47% memory usage during training, and supports up to 2x more training tokens - all while maintaining negligible impact on serving latency.

cs.LG↗

arXiv2024

OneAdapt: Fast Configuration Adaptation for Video Analytics Applications via Backpropagation

Deep learning inference on streaming media data, such as object detection in video or LiDAR feeds and text extraction from audio waves, is now ubiquitous. To achieve high inference accuracy, these applications typically require significant network bandwidth to gather high-fidelity data and extensive GPU resources to run deep neural networks (DNNs). While the high demand for network bandwidth and GPU resources could be substantially reduced by optimally adapting the configuration knobs, such as video resolution and frame rate, current adaptation techniques fail to meet three requirements simultaneously: adapt configurations (i) with minimum extra GPU or bandwidth overhead; (ii) to reach near-optimal decisions based on how the data affects the final DNN's accuracy, and (iii) do so for a range of configuration knobs. This paper presents OneAdapt, which meets these requirements by leveraging a gradient-ascent strategy to adapt configuration knobs. The key idea is to embrace DNNs' differentiability to quickly estimate the accuracy's gradient to each configuration knob, called AccGrad. Specifically, OneAdapt estimates AccGrad by multiplying two gradients: InputGrad (i.e. how each configuration knob affects the input to the DNN) and DNNGrad (i.e. how the DNN input affects the DNN inference output). We evaluate OneAdapt across five types of configurations, four analytic tasks, and five types of input data. Compared to state-of-the-art adaptation schemes, OneAdapt cuts bandwidth usage and GPU usage by 15-59% while maintaining comparable accuracy or improves accuracy by 1-5% while using equal or fewer resources.

cs.LG↗

arXiv2024

Cascade-Zero123: One Image to Highly Consistent 3D with Self-Prompted Nearby Views

Synthesizing multi-view 3D from one single image is a significant but challenging task. Zero-1-to-3 methods have achieved great success by lifting a 2D latent diffusion model to the 3D scope. The target view image is generated with a single-view source image and the camera pose as condition information. However, due to the high sparsity of the single input image, Zero-1-to-3 tends to produce geometry and appearance inconsistency across views, especially for complex objects. To tackle this issue, we propose to supply more condition information for the generation model but in a self-prompt way. A cascade framework is constructed with two Zero-1-to-3 models, named Cascade-Zero123, which progressively extract 3D information from the source image. Specifically, several nearby views are first generated by the first model and then fed into the second-stage model along with the source image as generation conditions. With amplified self-prompted condition images, our Cascade-Zero123 generates more consistent novel-view images than Zero-1-to-3. Experiment results demonstrate remarkable promotion, especially for various complex and challenging scenes, involving insects, humans, transparent objects, and stacked multiple objects etc. More demos and code are available at https://cascadezero123.github.io.

cs.CV↗

arXiv2024

BotEval: Facilitating Interactive Human Evaluation

Following the rapid progress in natural language processing (NLP) models, language models are applied to increasingly more complex interactive tasks such as negotiations and conversation moderations. Having human evaluators directly interact with these NLP models is essential for adequately evaluating the performance on such interactive tasks. We develop BotEval, an easily customizable, open-source, evaluation toolkit that focuses on enabling human-bot interactions as part of the evaluation process, as opposed to human evaluators making judgements for a static input. BotEval balances flexibility for customization and user-friendliness by providing templates for common use cases that span various degrees of complexity and built-in compatibility with popular crowdsourcing platforms. We showcase the numerous useful features of BotEval through a study that evaluates the performance of various chatbots on their effectiveness for conversational moderation and discuss how BotEval differs from other annotation tools.

cs.CL↗

arXiv2024

CacheGen: KV Cache Compression and Streaming for Fast Large Language Model Serving

As large language models (LLMs) take on complex tasks, their inputs are supplemented with longer contexts that incorporate domain knowledge. Yet using long contexts is challenging, as nothing can be generated until the whole context is processed by the LLM. While the context-processing delay can be reduced by reusing the KV cache of a context across different inputs, fetching the KV cache, which contains large tensors, over the network can cause high extra network delays. CacheGen is a fast context-loading module for LLM systems. First, CacheGen uses a custom tensor encoder, leveraging KV cache's distributional properties to encode a KV cache into more compact bitstream representations with negligible decoding overhead, to save bandwidth usage. Second, CacheGen adapts the compression level of different parts of a KV cache to cope with changes in available bandwidth, in order to maintain low context-loading delay and high generation quality. % When available bandwidth drops, CacheGen may raise the compression level for a part of the context or recompute its KV cache on the fly. We test CacheGen on popular LLMs and datasets. Compared to the recent systems that reuse the KV cache, CacheGen reduces the KV cache size by 3.5-4.3x and the total delay in fetching and processing contexts by 3.2-3.7x with negligible impact on the LLM response quality. Our code is at: https://github.com/UChi-JCL/CacheGen.

cs.NI↗

arXiv2024

Can Language Model Moderators Improve the Health of Online Discourse?

Conversational moderation of online communities is crucial to maintaining civility for a constructive environment, but it is challenging to scale and harmful to moderators. The inclusion of sophisticated natural language generation modules as a force multiplier to aid human moderators is a tantalizing prospect, but adequate evaluation approaches have so far been elusive. In this paper, we establish a systematic definition of conversational moderation effectiveness grounded on moderation literature and establish design criteria for conducting realistic yet safe evaluation. We then propose a comprehensive evaluation framework to assess models' moderation capabilities independently of human intervention. With our framework, we conduct the first known study of language models as conversational moderators, finding that appropriately prompted models that incorporate insights from social science can provide specific and fair feedback on toxic behavior but struggle to influence users to increase their levels of respect and cooperation.

cs.CL↗

arXiv2024

Task Aligned Meta-learning based Augmented Graph for Cold-Start Recommendation

The cold-start problem is a long-standing challenge in recommender systems due to the lack of user-item interactions, which significantly hurts the recommendation effect over new users and items. Recently, meta-learning based methods attempt to learn globally shared prior knowledge across all users, which can be rapidly adapted to new users and items with very few interactions. Though with significant performance improvement, the globally shared parameter may lead to local optimum. Besides, they are oblivious to the inherent information and feature interactions existing in the new users and items, which are critical in cold-start scenarios. In this paper, we propose a Task aligned Meta-learning based Augmented Graph (TMAG) to address cold-start recommendation. Specifically, a fine-grained task aligned constructor is proposed to cluster similar users and divide tasks for meta-learning, enabling consistent optimization direction. Besides, an augmented graph neural network with two graph enhanced approaches is designed to alleviate data sparsity and capture the high-order user-item interactions. We validate our approach on three real-world datasets in various cold-start scenarios, showing the superiority of TMAG over state-of-the-art methods for cold-start recommendation.

cs.IR↗

arXiv2024

Data/moment-driven approaches for fast predictive control of collective dynamics

Feedback control synthesis for large-scale particle systems is reviewed in the framework of model predictive control (MPC). The high-dimensional character of collective dynamics hampers the performance of traditional MPC algorithms based on fast online dynamic optimization at every time step. Two alternatives to MPC are proposed. First, the use of supervised learning techniques for the offline approximation of optimal feedback laws is discussed. Then, a procedure based on sequential linearization of the dynamics based on macroscopic quantities of the particle ensemble is reviewed. Both approaches circumvent the online solution of optimal control problems enabling fast, real-time, feedback synthesis for large-scale particle systems. Numerical experiments assess the performance of the proposed algorithms.

math.OC↗