arXiv subjects

Juncheng Yang

Publications and source records attributed to Juncheng Yang.

At least 19 recordsLinked to original sources

arXiv2026

TrajectoryDB: A New Database for Agent Trajectories

AI agents generate rich execution trajectories that capture their interactions with large language models, tools, and external environments. These trajectories are increasingly valuable for downstream tasks such as memory extraction, model fine-tuning, runtime optimization, and security and cost monitoring. Yet trajectory data today is fragmented across files, databases, and observability systems, with no persistent data management system designed around its unique structure and access patterns. We argue that trajectories should be treated as a distinct data type. A trajectory combines hierarchical execution structure, large volumes of text whose analysis often requires semantic reasoning, and rich dependencies and lineage among events, intermediate states, and derived artifacts. These properties introduce new requirements throughout the data lifecycle. Ingestion must reconstruct and preserve execution structure and lineage; storage must efficiently organize large but highly redundant contexts while maintaining relationships among records; and query processing must jointly reason over structure, temporal order, semantics, and lineage. We therefore envision TrajectoryDB, a trajectory-native data management system that co-designs ingestion, storage, and query processing to efficiently manage and analyze agent execution trajectories.

cs.DB↗

arXiv2026

HeadWiseKV: Budgeted Per-Head Cache Residency for Hybrid Long-Context Language Models

Long-context inference retains a growing key--value (KV) cache during decoding, which consumes substantial GPU memory and can reduce generation throughput. This bottleneck remains in hybrid language models because their residual global-attention layers can dominate context-dependent cache demand. We study how to allocate this state under an aggregate KV-residency budget. We introduce HeadWiseKV, a training-free framework that compresses the residual global KV caches of hybrid language models while preserving their native local, recurrent, and linear paths. It assigns each physical KV head a static, multilevel history window, making cache demand predictable before serving. We formulate this allocation as a restricted operational rate--distortion problem and propose SeqCalib as the core policy-generation algorithm in HeadWiseKV. SeqCalib processes layers in execution order and conditions each decision on the lower-layer policy used at deployment, thereby accounting for interactions across depth. A grouped-cache runtime materializes the selected policy as actual per-head KV residency rather than a mask over a full cache. We evaluate downstream quality across four hybrid long-context models and study physical residency and serving behavior on Qwen3.6-27B. HeadWiseKV retains near-Full-KV RULER and LoCoMo quality across the evaluated models. In the fixed-model systems study, it reduces sampled peak device memory by 8.59\% at a 112K context length and extends the largest verified successful context from 114K to 161K.

cs.AI↗

arXiv2026

Demystifying and Improving Lazy Promotion in Cache Eviction

Cache eviction algorithms play a critical role in the performance of modern data systems, yet their scalability is often limited by the high computational overhead associated with object promotions. Lazy Promotion techniques have emerged as relaxations of traditional Least-Recently-Used (LRU) methods, designed to alleviate lock contention and increase throughput. This work uses production traces from real-world systems to benchmark five Lazy Promotion strategies: Probabilistic-LRU, Batch-LRU, Delay-LRU, FIFO-reinsertion, and Random-LRU. We evaluate these techniques across miss ratio, scalability, promotion count, and a novel metric called promotion efficiency, which measures the number of hits per promotion. Our results reveal that Delay-LRU and FIFO-reinsertion significantly improve promotion efficiency, whereas Batch-LRU and Probabilistic-LRU struggle to reduce promotions without significantly increasing miss ratio. We further explore the impact of lazy promotion in advanced algorithms such as ARC and 2Q and make a similar observation. Moreover, we uncover substantial optimization potential, showing that most cache promotions are unnecessary when equipped with oracle knowledge. To further reduce promotions in LRU, we propose two novel enhancements-Delayed FIFO-reinsertion (D-FR) and Age-Guided Eviction (AGE)-that reduce promotions by 20-60% while achieving a similar or lower miss ratio.

cs.DB↗

arXiv2026

Learning-Augmented Heuristics: Simple, yet Smart, Robust and Interpretable Cache Eviction

Caching is widely used across the system stack to improve performance and efficiency, with eviction algorithms at its core. Existing cache eviction policies fall into two broad categories: static heuristics (e.g., 2Q, S3-FIFO) and smart algorithms (e.g., ARC, LRB). Smart caches can adapt to workloads and have the potential to achieve higher efficiency and robustness than static heuristics. However, we find that existing smart caches suffer from objective mismatches and instability. We introduce Learning-Augmented Heuristics (LAH), a framework that learns the cache-level parameters of static heuristics. By decoupling the data and control planes, LAH supports simple, high-speed data reads and writes on the data plane, while performing occasional asynchronous learning on the control plane using cache-level features. We demonstrate the effectiveness of LAH through S4-FIFO, a Smart S3-FIFO cache eviction algorithm. We pre-train a single model on 4,140 production traces and embed it in S4-FIFO to learn optimal cache parameters. On 1,035 evaluation traces, S4-FIFO improves the mean efficiency by 26% compared to S3-FIFO and by 8% compared to 3L-Cache, the best state-of-the-art algorithm. S4-FIFO is also robust---increasing miss ratio over FIFO by 0.8% on the worst trace, whereas 3L-Cache increases FIFO's miss ratio by 8.8%. Finally, S4-FIFO's decisions are also interpretable: a language model can provide a rationale for why a particular configuration was chosen.

cs.DC↗

arXiv2026

A Year in LLM Serving: Workload Evolution, Caching and Load-Balancing

Large Language Model (LLM) serving has become a critical cloud workload, and realistic traces are essential for motivating and benchmarking serving systems. However, existing LLM serving workload studies remain limited in scale and scope. They often observe short time periods and provide limited visibility into how users interact with models in production. As a result, they do not fully capture how LLM serving workloads evolve over time or how user-model interactions shape production traffic. In this work, we further the understanding of real-world LLM serving workloads through both a global characterization and a longitudinal study of a one-year production trace from CompanyX. Unlike prior studies, our trace captures full production behavior across many models and users, including both popular and long-tail models. We analyze the workload from aggregate, temporal, model-level, and user-level perspectives, revealing workload evolution and user-model structure that are typically hidden behind aggregate views. To support future research, we publicly release the full one-year trace, enabling downstream studies of production behavior without relying on sampled or synthetically generated workloads. The trace is available at https://github.com/HarvardMadSys/chutes_workload.

cs.AI↗

arXiv2026

LAANN: I/O-Aware Look-Ahead Search for Disk-Based Approximate Nearest Neighbor Search

Approximate nearest neighbor search (ANNS) is a fundamental primitive in large-scale retrieval, recommendation, and AI systems. As vector datasets grow to billions or even trillions of items, disk-based ANNS systems have emerged to handle this scale by storing vector data and index structures on storage systems, but their query performance remains dominated by I/O latency. Existing disk-based ANNS systems primarily optimize I/O efficiency or overlap I/O with computation, but they treat CPU computation and I/O access as largely separate components. This separation misses a critical opportunity: selectively processing candidates already cached in memory before making I/O decisions can reduce unnecessary disk accesses and improve search quality. However, exploiting this opportunity is challenging because excessive computation can delay critical I/O operations, while poorly chosen computation provides little benefit, potentially increasing overall query latency. In this paper, we present LAANN, a disk-based ANNS system that makes graph search explicitly I/O-aware by co-optimizing CPU computation and I/O access. LAANN combines three techniques: look-ahead search, which adapts the search strategy across query stages to balance I/O reduction and timely I/O issuance; a priority I/O-CPU pipeline, which uses I/O waiting time to process candidates cached in memory according to their expected impact on upcoming I/O decisions; and a fast lightweight in-memory graph index, which provides high-quality initial candidates to accelerate convergence and reduce disk accesses. Experiments on million- and billion-scale datasets demonstrate that LAANN substantially outperforms state-of-the-art disk-based ANNS systems. At Recall@10 = 0.9, LAANN achieves 1.41x-4.66x higher throughput, 29%-79% lower latency, and 1.59x-6.34x fewer I/O operations.

cs.DB↗

arXiv2026

LatentBox: Storing AI-Generated Images at Scale via a Latent-First Design

The explosive growth of AI-generated images has created a sustainability challenge for storage infrastructure. Platforms like Midjourney and Adobe Firefly already host billions of generative images, yet conventional object stores persist them as blobs with full-resolution pixels, consuming huge amounts of storage capacity and bandwidth. Unlike natural photos, however, AI-generated images can be deterministically reconstructed from compact, model-native latent tensors, making persistent image storage fundamentally redundant. This paper presents LatentBox, a latent-first storage system for AI-generated images. LatentBox treats compressed latents as durable storage objects and uses on-demand GPU reconstruction on the read path to trade inexpensive compute for large persistent storage savings. Our design is guided by the first large-scale analysis of AI-generated image access we are aware of, based on a 35-month, 2-billion-request production trace from a major generative-content platform. Motivated by the trace analysis, LatentBox keeps frequently accessed images in decoded pixel format for fast hits, stores less-active objects as compressed latents to expand effective cache capacity, and continuously adjusts the splits between the image and latent cache to optimize user-perceived access latency.We build a LatentBox prototype and evaluate it with the production trace. LatentBox reduces persistent storage by 78.7% with competitive or even lower mean and tail latency over a pure image-based storage.

cs.DC↗

arXiv2026

MoE-Prefill: Zero Redundancy Overheads in MoE Prefill Serving

Production LLM workloads increasingly serve discriminative tasks, such as classification, recommendation, and verification, whose answers are read from the logits of a single prefill pass with no autoregressive decoding. Serving these prefill-only workloads on mixture-of-experts (MoE) models is bottlenecked not by compute but by the distributed execution required to fit the model: existing parallel strategies (tensor, expert, and pipeline parallelism) trade memory pressure for redundant computation, communication, and synchronization, severely degrading MoE prefill serving efficiency. We observe that these overheads stem from coupling expert placement with synchronous activation routing -- a design inherited from the decoding era. The long, compute-bound forward passes of large-batch prefill open a per-layer window wide enough to stream expert weights in the background, replacing per-layer activation AllToAll with asynchronous weight AllGather fully overlapped with computation. We propose MoE-Prefill, a prefill-only serving system whose backend, AsyncEP (Asynchronous Expert Parallelism), gathers experts by weight rather than routing them by activation, and whose frontend co-enforces a physically-derived saturation threshold through prefix-aware routing and true-FLOPs load tracking. On Qwen3-235B-A22B across four hardware/precision configurations, MoE-Prefill delivers 1.35-1.37x throughput over the strongest distributed baseline on real-world workloads and up to 1.59x on long-context synthetic workloads, sustaining 29.8-36.2% per-GPU model FLOPs utilization.

cs.LG↗

arXiv2026

TStore: Rethinking AI Model Hub with Tensor-Centric Compression

Modern AI models are growing rapidly in size and redundancy, leading to significant storage and distribution challenges in model hubs. We present TStore, a tensor-centric system for reducing storage overhead through fine-grained deduplication and compression. TStore leverages tensor-level fingerprinting and clustering to identify redundancy across models without requiring annotations. Our design enables efficient storage reduction while preserving model usability and performance. Experiments on real-world model repositories demonstrate substantial storage savings with minimal overhead.

cs.DC↗

arXiv2025

Clock2Q+: A Simple and Efficient Replacement Algorithm for Metadata Cache in VMware vSAN

Cache replacement algorithms are critical building blocks of storage systems. This paper examines the characteristics of metadata caches and argues that they inherently exhibit correlated references, even when the corresponding data accesses do not contain correlated references. The presence of correlated references reduces the effectiveness of cache replacement algorithms because these references are often mistakenly categorized as hot blocks. Clock2Q+ is specifically designed for metadata caches and has been implemented in vSAN and VDFS, two flagship storage products of VMware by Broadcom. Similar to S3-FIFO, Clock2Q+ uses three queues; however, Clock2Q+ introduces a correlation window in the Small FIFO queue, where blocks in this window do not set the reference bit. This simple enhancement allows Clock2Q+ to outperform state-of-the-art replacement algorithms. Compared to S3-FIFO, the second-best performing algorithm, Clock2Q+ achieves up to a 28.5% lower miss ratio on metadata traces. Clock2Q+ possesses the essential properties required for large-scale storage systems: it has low CPU overhead on cache hits, low memory overhead, scales efficiently to multiple CPUs, and is both easy to tune and implement. Additionally, Clock2Q+ outperforms state-of-the-art cache replacement algorithms on data traces as well.

cs.DC↗

arXiv2025

Arctic Inference with Shift Parallelism: Fast and Efficient Open Source Inference System for Enterprise AI

Inference is now the dominant AI workload, yet existing systems force trade-offs between latency, throughput, and cost. Arctic Inference, an open-source vLLM plugin from Snowflake AI Research, introduces Shift Parallelism, a dynamic parallelism strategy that adapts to real-world traffic while integrating speculative decoding, SwiftKV compute reduction, and optimized embedding inference. It achieves up to 3.4 times faster request completion, 1.75 times faster generation, and 1.6M tokens/sec per GPU for embeddings, outperforming both latency- and throughput-optimized deployments. Already powering Snowflake Cortex AI, Arctic Inference delivers state-of-the-art, cost-effective inference for enterprise AI and is now available to the community.

cs.DC↗

arXiv2025

MorphServe: Efficient and Workload-Aware LLM Serving via Runtime Quantized Layer Swapping and KV Cache Resizing

Efficiently serving large language models (LLMs) under dynamic and bursty workloads remains a key challenge for real-world deployment. Existing serving frameworks and static model compression techniques fail to adapt to workload fluctuations, leading to either service-level objective (SLO) violations under full-precision serving or persistent accuracy degradation with static quantization. We present MorphServe, a dynamic, workload-aware LLM serving framework based on morphological adaptation. MorphServe introduces two asynchronous, token-level runtime mechanisms: quantized layer swapping, which selectively replaces less impactful layers with quantized alternatives during high-load periods, and pressure-aware KV cache resizing, which dynamically adjusts KV cache capacity in response to memory pressure. These mechanisms enable state-preserving transitions with minimum runtime overhead and are fully compatible with modern scheduling and attention techniques. Extensive experiments on Vicuna and Llama family models with real-world workloads demonstrate that MorphServe reduces average SLO violations by 92.45 percent and improves the P95 TTFT latency by 2.2x-3.9x compared to full-precision serving, without compromising generation quality. These results establish MorphServe as a practical and elastic solution for LLM deployment in dynamic environments.

cs.DC↗

arXiv2025

ZipLLM: Efficient LLM Storage via Model-Aware Synergistic Data Deduplication and Compression

Modern model hubs, such as Hugging Face, store tens of petabytes of LLMs, with fine-tuned variants vastly outnumbering base models and dominating storage consumption. Existing storage reduction techniques -- such as deduplication and compression -- are either LLM-oblivious or not compatible with each other, limiting data reduction effectiveness. Our large-scale characterization study across all publicly available Hugging Face LLM repositories reveals several key insights: (1) fine-tuned models within the same family exhibit highly structured, sparse parameter differences suitable for delta compression; (2) bitwise similarity enables LLM family clustering; and (3) tensor-level deduplication is better aligned with model storage workloads, achieving high data reduction with low metadata overhead. Building on these insights, we design BitX, an effective, fast, lossless delta compression algorithm that compresses XORed difference between fine-tuned and base LLMs. We build ZipLLM, a model storage reduction pipeline that unifies tensor-level deduplication and lossless BitX compression. By synergizing deduplication and compression around LLM family clustering, ZipLLM reduces model storage consumption by 54%, over 20% higher than state-of-the-art deduplication and compression approaches.

cs.DB↗

arXiv2024

Helix: Serving Large Language Models over Heterogeneous GPUs and Network via Max-Flow

This paper introduces Helix, a distributed system for high-throughput, low-latency large language model (LLM) serving in heterogeneous GPU clusters. The key idea behind Helix is to formulate inference computation of LLMs over heterogeneous GPUs and network connections as a max-flow problem on directed, weighted graphs, whose nodes represent GPU instances and edges capture both GPU and network heterogeneity through their capacities. Helix then uses a mixed integer linear programming (MILP) algorithm to discover highly optimized strategies to serve LLMs on heterogeneous GPUs. This approach allows Helix to jointly optimize model placement and request scheduling, two highly entangled tasks in heterogeneous LLM serving. Our evaluation on several heterogeneous clusters ranging from 24 to 42 GPU nodes shows that Helix improves serving throughput by up to 3.3x and reduces prompting and decoding latency by up to 66% and 24%, respectively, compared to existing approaches. Helix is available at https://github.com/Thesys-lab/Helix-ASPLOS25.

cs.DC↗

arXiv2024

Can Increasing the Hit Ratio Hurt Cache Throughput? (Long Version)

Software caches are an intrinsic component of almost every computer system. Consequently, caching algorithms, particularly eviction policies, are the topic of many papers. Almost all these prior papers evaluate the caching algorithm based on its hit ratio, namely the fraction of requests that are found in the cache, as opposed to disk. The hit ratio is viewed as a proxy for traditional performance metrics like system throughput or response time. Intuitively it makes sense that higher hit ratio should lead to higher throughput (and lower response time), since more requests are found in the cache (low access time) as opposed to the disk (high access time). This paper challenges this intuition. We show that increasing the hit ratio can actually hurt the throughput (and response time) for many caching algorithms. Our investigation follows a three-pronged approach involving (i) queueing modeling and analysis, (ii) implementation and measurement, and (iii) simulation to validate the accuracy of the queueing model. We also show that the phenomenon of throughput decreasing at higher hit ratios is likely to be more pronounced in future systems, where the trend is towards faster disks and higher numbers of cores per CPU.

cs.PF↗

arXiv2024

DSDRNet: Disentangling Representation and Reconstruct Network for Domain Generalization

Domain generalization faces challenges due to the distribution shift between training and testing sets, and the presence of unseen target domains. Common solutions include domain alignment, meta-learning, data augmentation, or ensemble learning, all of which rely on domain labels or domain adversarial techniques. In this paper, we propose a Dual-Stream Separation and Reconstruction Network, dubbed DSDRNet. It is a disentanglement-reconstruction approach that integrates features of both inter-instance and intra-instance through dual-stream fusion. The method introduces novel supervised signals by combining inter-instance semantic distance and intra-instance similarity. Incorporating Adaptive Instance Normalization (AdaIN) into a two-stage cyclic reconstruction process enhances self-disentangled reconstruction signals to facilitate model convergence. Extensive experiments on four benchmark datasets demonstrate that DSDRNet outperforms other popular methods in terms of domain generalization capabilities.

cs.CV↗

arXiv2024

Cross-Modal Adapter: Parameter-Efficient Transfer Learning Approach for Vision-Language Models

Adapter-based parameter-efficient transfer learning has achieved exciting results in vision-language models. Traditional adapter methods often require training or fine-tuning, facing challenges such as insufficient samples or resource limitations. While some methods overcome the need for training by leveraging image modality cache and retrieval, they overlook the text modality's importance and cross-modal cues for the efficient adaptation of parameters in visual-language models. This work introduces a cross-modal parameter-efficient approach named XMAdapter. XMAdapter establishes cache models for both text and image modalities. It then leverages retrieval through visual-language bimodal information to gather clues for inference. By dynamically adjusting the affinity ratio, it achieves cross-modal fusion, decoupling different modal similarities to assess their respective contributions. Additionally, it explores hard samples based on differences in cross-modal affinity and enhances model performance through adaptive adjustment of sample learning intensity. Extensive experimental results on benchmark datasets demonstrate that XMAdapter outperforms previous adapter-based methods significantly regarding accuracy, generalization, and efficiency.

cs.CV↗

arXiv2024

Soft-Prompting with Graph-of-Thought for Multi-modal Representation Learning

The chain-of-thought technique has been received well in multi-modal tasks. It is a step-by-step linear reasoning process that adjusts the length of the chain to improve the performance of generated prompts. However, human thought processes are predominantly non-linear, as they encompass multiple aspects simultaneously and employ dynamic adjustment and updating mechanisms. Therefore, we propose a novel Aggregation-Graph-of-Thought (AGoT) mechanism for soft-prompt tuning in multi-modal representation learning. The proposed AGoT models the human thought process not only as a chain but also models each step as a reasoning aggregation graph to cope with the overlooked multiple aspects of thinking in single-step reasoning. This turns the entire reasoning process into prompt aggregation and prompt flow operations. Experiments show that our multi-modal model enhanced with AGoT soft-prompting achieves good results in several tasks such as text-image retrieval, visual question answering, and image recognition. In addition, we demonstrate that it has good domain generalization performance due to better reasoning.

cs.AI↗