arXiv subjects

Shouqian Shi

Publications and source records attributed to Shouqian Shi.

8 recordsLinked to original sources

arXiv2026

IRIS: Reusable Identity Representations from Frozen LLMs for Entity Alignment

Entity alignment (EA) identifies entities across knowledge graphs (KGs) that refer to the same real-world object. Conventional EA methods mainly exploit explicit graph structures and textual fields, which often provide insufficient semantic understanding to recognize the same entity under heterogeneous descriptions and distinguish it from semantically similar entities. Although large language models (LLMs) offer deeper entity understanding, existing LLM-based EA methods largely use this capability for auxiliary generation or candidate-conditioned decisions. Consequently, such understanding is not distilled into a stable and directly comparable identity space, leaving alignment tied to specific KG pairs or candidate sets and requiring repeated processing as the matching context changes. To address these limitations, we propose IRIS (Identity Representations from Internal States), a training-free framework that constructs for each entity an iris-like signature encoding its distinctive and stable identity characteristics. IRIS derives these signatures by eliciting identity-oriented contextual representations from a frozen LLM, thereby forming a shared space in which each entity is encoded once and can be aligned across different KGs through direct similarity comparison, without pair-dependent representation construction or candidate-wise LLM inference. Across four established EA benchmarks and two frozen LLM backbones, the best IRIS variants achieve Hits@1 scores of 100.00, 99.38, 98.31, and 97.99 on D-Y-15K V2, DBP-WIKI, ICEWS-WIKI, and ICEWS-YAGO, respectively.

cs.CL↗

arXiv2026

Penelope: Localized Latent Recurrence for Efficient Structured Reasoning

Complex structured reasoning tasks often require additional computation, yet current language models obtain it mainly by increasing parameter scale or by serializing intermediate steps as chain-of-thought (CoT) tokens. The former raises training and deployment costs, while the latter ties reasoning computation to autoregressive output length. We introduce Penelope, an efficient latent-reasoning framework for pretrained decoder-only Transformers that localizes recurrent computation to a selected decoder interval. The lower decoder prefix is evaluated once to construct a problem-conditioned boundary memory, which is then iteratively refined through time-modulated GRU dynamics and recurrent readout states before answer generation. A progressive CoT-to-latent curriculum transfers visible reasoning into this internal recurrent path, allowing additional computation to be allocated in latent space without repeatedly executing the complete decoder or generating a long intermediate trace. Experiments on open-source structured-reasoning benchmarks show that, at validation-selected latent budgets, Penelope attains competitive accuracy relative to established latent-reasoning models while reducing measured inference latency. These results show that latent refinement can be localized to a narrow decoder interval, reducing repeated full-decoder execution without generating a long visible reasoning trace and providing a practical accuracy-efficiency tradeoff for decoder-only Transformer models.

cs.AI↗

arXiv2026

TraceCLIP: Recovering Local Semantics from Patch-to-CLS Contributions

Dense vision-language understanding, including object localization, region recognition, and open-vocabulary semantic segmentation, requires associating language concepts with spatially grounded visual regions. CLIP provides a strong foundation for these tasks by learning a shared image-text embedding space from large-scale contrastive pre-training. However, its image-level objective aligns text with a CLS-derived global representation, leaving local vision-language correspondence only indirectly constrained. Existing methods either introduce additional supervision, external models, or task-specific adaptation, while training-free approaches mainly recover dense responses from existing patch features without examining where local semantics become most accessible within CLIP. We introduce TraceCLIP, a training-free framework that recovers latent patch-level semantic evidence by isolating the patch-specific terms written into the CLS attention output. TraceCLIP further converts contribution-derived semantic responses into a semantic-geodesic topology gate that calibrates final-layer patch affinity for dense feature reconstruction. Diagnostic experiments show that these contribution features exhibit strong local semantic discrimination and text-conditioned spatial alignment. On eight zero-shot semantic segmentation benchmarks, TraceCLIP achieves gains of 1.3 to 4.5 points in average mIoU over the strongest prior training-free methods across both backbones and background settings, without additional training, external vision foundation models, or region-level supervision. More broadly, these findings suggest that spatially localized semantics may remain accessible within the internal construction of globally aligned representations.

cs.CV↗

arXiv2026

Lens: Bringing the Right Semantic Perspective into Focus for Training-Free Multimodal Representation Learning

High-quality representations are essential for a wide range of downstream tasks. Dedicated embedding models are explicitly optimized for representation learning, yet their training data are often more limited in scale and diversity than the massive corpora used to pretrain modern large language models and multimodal large language models. Large-scale pretraining and instruction following enable autoregressive models to select relevant evidence, integrate multimodal information, and infer semantics under different task perspectives, creating a distinctive opportunity for training-free representation learning. However, our analysis reveals that existing semantic-elicitation methods do not reliably orient the extracted states toward the semantic perspective required by the downstream task. Consequently, the resulting representations often remain dominated by salient input content. We characterize this problem as semantic perspective misalignment and propose Lens, a training-free framework that makes representation readout task-directed. Semantic Perspective Anchoring associates the task-required perspective with a task-specific readout phrase, specifying the interpretive role of the positions later used for extraction. Contextualized Phrase Readout places the same phrase after the complete input and aggregates its token states, combining full-context access with the anchored perspective. The resulting representation reflects task-conditioned evidence integration and inference rather than a generic summary of salient content. Without parameter updates, architectural modification, or reranking, Lens achieves an overall Precision@1 of 63.9 across all 36 MMEB datasets, outperforming the closest same-backbone training-free embedding baseline by 10.2 points.

cs.CL↗

arXiv2025

Outback: Fast and Communication-efficient Index for Key-Value Store on Disaggregated Memory

Disaggregated memory systems achieve resource utilization efficiency and system scalability by distributing computation and memory resources into distinct pools of nodes. RDMA is an attractive solution to support high-throughput communication between different disaggregated resource pools. However, existing RDMA solutions face a dilemma: one-sided RDMA completely bypasses computation at memory nodes, but its communication takes multiple round trips; two-sided RDMA achieves one-round-trip communication but requires non-trivial computation for index lookups at memory nodes, which violates the principle of disaggregated memory. This work presents Outback, a novel indexing solution for key-value stores with a one-round-trip RDMA-based network that does not incur computation-heavy tasks at memory nodes. Outback is the first to utilize dynamic minimal perfect hashing and separates its index into two components: one memory-efficient and compute-heavy component at compute nodes and the other memory-heavy and compute-efficient component at memory nodes. We implement a prototype of Outback and evaluate its performance in a public cloud. The experimental results show that Outback achieves higher throughput than both the state-of-the-art one-sided RDMA and two-sided RDMA-based in-memory KVS by 1.06-5.03x, due to the unique strength of applying a separated perfect hashing index.

cs.DB↗

arXiv2021

WebFlow: Scalable and Decentralized Routing for Payment Channel Networks with High Resource Utilization

Payment channel networks (PCNs) have been designed and utilized to address the scalability challenge and throughput limitation of blockchains. Routing is a core problem of PCNs. An ideal PCN routing method needs to achieve 1) high scalability that can maintain low per-node memory and communication cost for large PCNs, 2) high resource utilization of payment channels, and 3) the privacy of users. However, none of the existing PCN systems consider all these requirements. In this work, we propose WebFlow, a distributed routing solution for PCNs, which only requires each user to maintain localized information and can be used for massive-scale networks with high resource utilization. We make use of two distributed data structures: multi-hop Delaunay triangulation (MDT) originally proposed for wireless networks and our innovation called distributed Voronoi diagram. We propose new protocols to generate a virtual Euclidean space in order to apply MDT to PCNs and use the distributed Voronoi diagram to enhance routing privacy. We conduct extensive simulations and prototype implementation to further evaluate WebFlow. The results using real and synthetic PCN topologies and transaction traces show that WebFlow can achieve extremely low per-node overhead and a high success rate compared to existing methods.

cs.NI↗

arXiv2019

Modeling and Designing Routing Protocols in Quantum Networks

Quantum networks enable a number of important applications such as quantum key distribution. The basic function of a quantum network is to enable long-distance quantum entanglement between two remote communication parties. This work focuses on the entanglement routing problem, whose objective is to build long-distance entanglements for the concurrent source-destination pairs through multiple hops. Different from existing works that analyzes the traditional routing techniques on special network topologies, we present a comprehensive entanglement routing model that reflects the differences between quantum networks and classical networks and new entanglement routing algorithms that utilize the unique properties of quantum networks. Evaluation results show that the proposed algorithm Q-CAST increases the number of successful long-distance entanglements by a big margin compared to other methods. The model and simulator developed by this work may encourage more network researchers to study the entanglement routing problem.

cs.NI↗

arXiv2019

Concury: A Fast and Light-weighted Software Load Balancer

A load balancer (LB) is a vital network function for cloud services to balance the load amongst resources. Stateful software LBs that run on commodity servers provides flexibility, cost-efficiency, and packet consistency. However current designs have two main limitations: 1) states are stored as digests which may cause packet inconsistency due to digest collisions; 2) the data plane needs to update for every new connection, and frequent updates hurt throughput and packet consistency. In this work, we present a new software stateful LB called Concury, which is the first solution to solve these problems. The key innovation of Concury is an algorithmic approach to store and look up large network states with frequent connection arrivals, which is succinct in memory cost, consistent under network changes, and incurs infrequent data plane updates. The evaluation results show that the Concury algorithm provides 4x throughput and consumes less memory compared to other LB algorithms, while providing weighted load balancing and false-hit freedom, for both real and synthetic data center traffic. We implement Concury as a prototype system deployed in CloudLab and show that the throughput of Concury on a single thread can reach 62.5% of the maximum capacity of two 10GbE NICs and that on two threads can reach the maximum capacity.

cs.NI↗