arXiv subjects

Runhui Wang

Publications and source records attributed to Runhui Wang.

10 recordsLinked to original sources

arXiv2026

Cliff: Learning Process Rewards from the First Mistake

Reinforcement learning with verifiable rewards (RLVR) has emerged as a powerful paradigm for large language model (LLM) post-training, but its reliance on coarse outcome rewards leads to limited guidance on intermediate reasoning processes. Existing approaches such as process reward modeling and on-policy distillation introduce additional constraints, such as reliance on a specialized reward model or assuming identical reasoning patterns between teacher and student. Nevertheless, we observe that once a reasoning process first goes wrong, evaluating the subsequent reasoning provides limited additional information, as it is already conditioned on an invalid prefix. Therefore, we propose Cliff, a reward shaping strategy that utilizes an off-the-shelf LLM as a teacher to identify the first mistake in each rollout. As a result, the rollout is naturally decomposed into two parts: a correct prefix and an incorrect suffix. Cliff then converts this signal into token-level advantages, assigning positive advantages for the correct prefix and negative feedback afterward. Experiments across 12 different scenarios demonstrate that Cliff consistently improves reasoning performance, outperforming on-policy distillation by 15% and standard GRPO by 7%, even with teachers of modest capability. Furthermore, we analyse the role of ``ground truth'' in Cliff and investigate its training dynamics. These results establish Cliff as a simple, general and effective approach for improving RLVR with richer, fine-grained supervision.

cs.LG↗

arXiv2026

SkillsBench: Benchmarking How Well Agent Skills Work Across Diverse Tasks

Agent Skills are structured packages of procedural knowledge that augment large language model (LLM) agents at inference time. Despite rapid adoption, there is no standard way to measure whether they actually help. We present SkillsBench, a benchmark whose current inventory contains 87 tasks across 8 domains paired with curated Skills and deterministic verifiers. Our latest aggregate evaluation runs the 87-task benchmark under matched no-Skills and curated-Skills conditions for 18 model-harness configurations. Curated Skills raise the average pass rate from 33.9% to 50.5% (+16.6 percentage points; 25.5% normalized gain), with configuration-level gains ranging from +4.1 to +25.7 pp. Focused Skills with at most three modules outperform larger or exhaustive bundles, and smaller models with Skills can match larger models without them. SkillsBench establishes paired evaluation as the foundation for rigorous measurement of Skill efficacy on agentic, expertise-heavy work.

cs.AI↗

arXiv2025

Declarative Data Pipeline for Large Scale ML Services

Modern distributed data processing systems struggle to balance performance, maintainability, and developer productivity when integrating machine learning at scale. These challenges intensify in large collaborative environments due to high communication overhead and coordination complexity. We present a "Declarative Data Pipeline" (DDP) architecture that addresses these challenges while processing billions of records efficiently. Our modular framework seamlessly integrates machine learning within Apache Spark using logical computation units called Pipes, departing from traditional microservice approaches. By establishing clear component boundaries and standardized interfaces, we achieve modularity and optimization without sacrificing maintainability. Enterprise case studies demonstrate substantial improvements: 50% better development efficiency, collaboration efforts compressed from weeks to days, 500x scalability improvement, and 10x throughput gains.

cs.DC↗

arXiv2025

CSR-Bench: Benchmarking LLM Agents in Deployment of Computer Science Research Repositories

The increasing complexity of computer science research projects demands more effective tools for deploying code repositories. Large Language Models (LLMs), such as Anthropic Claude and Meta Llama, have demonstrated significant advancements across various fields of computer science research, including the automation of diverse software engineering tasks. To evaluate the effectiveness of LLMs in handling complex code development tasks of research projects, particularly for NLP/CV/AI/ML/DM topics, we introduce CSR-Bench, a benchmark for Computer Science Research projects. This benchmark assesses LLMs from various aspects including accuracy, efficiency, and deployment script quality, aiming to explore their potential in conducting computer science research autonomously. We also introduce a novel framework, CSR-Agents, that utilizes multiple LLM agents to automate the deployment of GitHub code repositories of computer science research projects. Specifically, by checking instructions from markdown files and interpreting repository structures, the model generates and iteratively improves bash commands that set up the experimental environments and deploy the code to conduct research tasks. Preliminary results from CSR-Bench indicate that LLM agents can significantly enhance the workflow of repository deployment, thereby boosting developer productivity and improving the management of developmental workflows.

cs.SE↗

arXiv2024

Learning from Natural Language Explanations for Generalizable Entity Matching

Entity matching is the task of linking records from different sources that refer to the same real-world entity. Past work has primarily treated entity linking as a standard supervised learning problem. However, supervised entity matching models often do not generalize well to new data, and collecting exhaustive labeled training data is often cost prohibitive. Further, recent efforts have adopted LLMs for this task in few/zero-shot settings, exploiting their general knowledge. But LLMs are prohibitively expensive for performing inference at scale for real-world entity matching tasks. As an efficient alternative, we re-cast entity matching as a conditional generation task as opposed to binary classification. This enables us to "distill" LLM reasoning into smaller entity matching models via natural language explanations. This approach achieves strong performance, especially on out-of-domain generalization tests (10.85% F-1) where standalone generative methods struggle. We perform ablations that highlight the importance of explanations, both for performance and model robustness.

cs.CL↗

arXiv2024

GRAM: Generative Retrieval Augmented Matching of Data Schemas in the Context of Data Security

Schema matching constitutes a pivotal phase in the data ingestion process for contemporary database systems. Its objective is to discern pairwise similarities between two sets of attributes, each associated with a distinct data table. This challenge emerges at the initial stages of data analytics, such as when incorporating a third-party table into existing databases to inform business insights. Given its significance in the realm of database systems, schema matching has been under investigation since the 2000s. This study revisits this foundational problem within the context of large language models. Adhering to increasingly stringent data security policies, our focus lies on the zero-shot and few-shot scenarios: the model should analyze only a minimal amount of customer data to execute the matching task, contrasting with the conventional approach of scrutinizing the entire data table. We emphasize that the zero-shot or few-shot assumption is imperative to safeguard the identity and privacy of customer data, even at the potential cost of accuracy. The capability to accurately match attributes under such stringent requirements distinguishes our work from previous literature in this domain.

cs.DB↗

arXiv2024

Neural Locality Sensitive Hashing for Entity Blocking

Locality-sensitive hashing (LSH) is a fundamental algorithmic technique widely employed in large-scale data processing applications, such as nearest-neighbor search, entity resolution, and clustering. However, its applicability in some real-world scenarios is limited due to the need for careful design of hashing functions that align with specific metrics. Existing LSH-based Entity Blocking solutions primarily rely on generic similarity metrics such as Jaccard similarity, whereas practical use cases often demand complex and customized similarity rules surpassing the capabilities of generic similarity metrics. Consequently, designing LSH functions for these customized similarity rules presents considerable challenges. In this research, we propose a neuralization approach to enhance locality-sensitive hashing by training deep neural networks to serve as hashing functions for complex metrics. We assess the effectiveness of this approach within the context of the entity resolution problem, which frequently involves the use of task-specific metrics in real-world applications. Specifically, we introduce NLSHBlock (Neural-LSH Block), a novel blocking methodology that leverages pre-trained language models, fine-tuned with a novel LSH-based loss function. Through extensive evaluations conducted on a diverse range of real-world datasets, we demonstrate the superiority of NLSHBlock over existing methods, exhibiting significant performance improvements. Furthermore, we showcase the efficacy of NLSHBlock in enhancing the performance of the entity matching phase, particularly within the semi-supervised setting.

cs.IR↗

arXiv2023

Language is All a Graph Needs

The emergence of large-scale pre-trained language models has revolutionized various AI research domains. Transformers-based Large Language Models (LLMs) have gradually replaced CNNs and RNNs to unify fields of computer vision and natural language processing. Compared with independent data samples such as images, videos or texts, graphs usually contain rich structural and relational information. Meanwhile, language, especially natural language, being one of the most expressive mediums, excels in describing complex structures. However, existing work on incorporating graph problems into the generative language modeling framework remains very limited. Considering the rising prominence of LLMs, it becomes essential to explore whether LLMs can also replace GNNs as the foundation model for graphs. In this paper, we propose InstructGLM (Instruction-finetuned Graph Language Model) with highly scalable prompts based on natural language instructions. We use natural language to describe multi-scale geometric structure of the graph and then instruction finetune an LLM to perform graph tasks, which enables Generative Graph Learning. Our method surpasses all GNN baselines on ogbn-arxiv, Cora and PubMed datasets, underscoring its effectiveness and sheds light on generative LLMs as new foundation model for graph machine learning. Our code is open-sourced at https://github.com/agiresearch/InstructGLM.

cs.CL↗

arXiv2022

Sudowoodo: Contrastive Self-supervised Learning for Multi-purpose Data Integration and Preparation

Machine learning (ML) is playing an increasingly important role in data management tasks, particularly in Data Integration and Preparation (DI&P). The success of ML-based approaches, however, heavily relies on the availability of large-scale, high-quality labeled datasets for different tasks. Moreover, the wide variety of DI&P tasks and pipelines oftentimes requires customizing ML solutions which can incur a significant cost for model engineering and experimentation. These factors inevitably hold back the adoption of ML-based approaches to new domains and tasks. In this paper, we propose Sudowoodo, a multi-purpose DI&P framework based on contrastive representation learning. Sudowoodo features a unified, matching-based problem definition capturing a wide range of DI&P tasks including Entity Matching (EM) in data integration, error correction in data cleaning, semantic type detection in data discovery, and more. Contrastive learning enables Sudowoodo to learn similarity-aware data representations from a large corpus of data items (e.g., entity entries, table columns) without using any labels. The learned representations can later be either directly used or facilitate fine-tuning with only a few labels to support different DI&P tasks. Our experiment results show that Sudowoodo achieves multiple state-of-the-art results on different levels of supervision and outperforms previous best specialized blocking or matching solutions for EM. Sudowoodo also achieves promising results in data cleaning and semantic type detection tasks showing its versatility in DI&P applications.

cs.DB↗

arXiv2019

Efficient Algorithms for Approximate Single-Source Personalized PageRank Queries

Given a graph $G$, a source node $s$ and a target node $t$, the personalized PageRank (PPR) of $t$ with respect to $s$ is the probability that a random walk starting from $s$ terminates at $t$. An important variant of the PPR query is single-source PPR (SSPPR), which enumerates all nodes in $G$, and returns the top-$k$ nodes with the highest PPR values with respect to a given source $s$. PPR in general and SSPPR in particular have important applications in web search and social networks, e.g., in Twitter's Who-To-Follow recommendation service. However, PPR computation is known to be expensive on large graphs, and resistant to indexing. Consequently, previous solutions either use heuristics, which do not guarantee result quality, or rely on the strong computing power of modern data centers, which is costly. Motivated by this, we propose effective index-free and index-based algorithms for approximate PPR processing, with rigorous guarantees on result quality. We first present FORA, an approximate SSPPR solution that combines two existing methods Forward Push (which is fast but does not guarantee quality) and Monte Carlo Random Walk (accurate but slow) in a simple and yet non-trivial way, leading to both high accuracy and efficiency. Further, FORA includes a simple and effective indexing scheme, as well as a module for top-$k$ selection with high pruning power. Extensive experiments demonstrate that the proposed solutions are orders of magnitude more efficient than their respective competitors. Notably, on a billion-edge Twitter dataset, FORA answers a top-500 approximate SSPPR query within 1 second, using a single commodity server.

cs.SI↗