Evaluating the coding capabilities of models through algorithmic code generation is challenging, as it requires deep problem understanding and complex algorithm design. Current benchmarks suffer from a narrow focus on final execution results (such as pass@k), neglecting the crucial reasoning and problem-solving processes inherent in code generation. To address this limitation, we introduce a multi-phase algorithmic code generation benchmark, MUPA, structured around human computational thinking. MUPA dissects the evaluation into four distinct phases: example understanding, algorithm selection, solution description, and code generation. This framework facilitates a comprehensive assessment by providing insights into the model's intermediate problem-solving steps, rather than just the final code. We manually curated 197 high-quality competitive programming problems from Codeforces. Utilizing an LLM-as-a-judge paradigm with specialized prompts, our rigorous evaluation of several existing code generation LLMs reveals significant across-the-board challenges. Notably, we establish a positive correlation, indicating that proficiency in an earlier phase directly impacts performance in subsequent phases, underscoring the interdependency of these algorithmic skills. The benchmark is publicly available at https://github.com/cheniison/MUPA.

Recent success of text-to-image (T2I) generation and its increasing practical applications, enabled by diffusion models, require urgent consideration of erasing unwanted concepts, e.g., copyrighted, offensive, and unsafe ones, from the pre-trained models in a precise, timely, and low-cost manner. The twofold demand of concept erasure includes not only a precise removal of the target concept (i.e., erasure efficacy) but also a minimal change on non-target content (i.e., prior preservation), during generation. Existing methods face challenges in maintaining an effective balance between erasure efficacy and prior preservation, and they can be computationally costly. To improve, we propose a precise, fast, and low-cost concept erasure method, called Adaptive Vaule Decomposer (AdaVD), which is training-free. Our method is grounded in a classical linear algebraic operation of computing orthogonal complement, implemented in the value space of each cross-attention layer within the UNet of diffusion models. We design a shift factor to adaptively navigate the erasure strength, enhancing effectively prior preservation without sacrificing erasure efficacy. Extensive comparative experiments with both training-based and training-free state of the arts demonstrate that the proposed AdaVD excels in both single and multiple concept erasure, showing 2 to 10 times of improvement in prior preservation than the second best, meanwhile achieving the best or near best erasure efficacy. AdaVD supports a series of diffusion models and downstream image generation tasks, with code available on: https://github.com/WYuan1001/AdaVD.

Recently, character-word lattice structures have achieved promising results for Chinese named entity recognition (NER), reducing word segmentation errors and increasing word boundary information for character sequences. However, constructing the lattice structure is complex and time-consuming, thus these lattice-based models usually suffer from low inference speed. Moreover, the quality of the lexicon affects the accuracy of the NER model. Since noise words can potentially confuse NER, limited coverage of the lexicon can cause lattice-based models to degenerate into partial character-based models. In this article, we propose a hierarchical label-enhanced contrastive learning (HLCL) method for Chinese NER. Instead of relying on the lattice structure, HLCL offers an alternative solution to robustly integrate entity boundary and type information with the help of both labels semantic and contrastive learning. HLCL is empowered by two techniques: 1) sentence-level contrastive learning (SCL) to model global mutual information between two different modalities (e.g., labels and sentences) and 2) token-level contrastive learning (TCL) to close the gap between representations of different characters (e.g., label-enhanced characters and original characters), resulting in local mutual information. With the well-designed contrastive learning scheme and the concise model during inference, HLCL can fully leverage the transferable label semantic and has a superb speed of inference. Experiments on four Chinese NER datasets show that HLCL obtains excellent efficiency as well as performance compared with existing lattice-based approaches.

The pervasive threat of shilling attacks, where malicious users inject fraudulent ratings to manipulate recommendations, critically undermines the reliability of Matrix Factorization (MF)-based recommender systems. This paper proposes AttResRec, a novel MF-based approach designed to improve system integrity by learning and integrating user credibility directly into the recommendation pipeline. AttResRec's defense is built upon three synergistic innovations. First, it employs a user credibility estimation mechanism that quantifies user credibility by assessing the consistency between an individual's interaction history and prevalent item co-occurrence patterns identified from collective user behavior. This directly enables differentiation between genuine and potentially malicious users. Second, the learned credibility dynamically informs a Credibility-aware Huber Loss (CHL) function. The CHL adaptively modifies its error sensitivity, rigorously penalizing deviations for high-credibility users while robustly limiting the influence of large errors associate with low-credibility users. Third, the model optimization is performed via Credibility-Weighted Stochastic Gradient Descent (CW-SGD), ensuring that users with lower credibility scores exert a diminished influence on the learned model parameters. Extensive experiments on the MovieLens-25M and Amazon Musical Instruments datasets, under diverse shilling attack scenarios, demonstrate AttResRec's benefits. That is, it not only achieves superior recommendation accuracy but also exhibits enhanced attack resistance, evidenced by lower prediction shift and hit ratios for poisoned items in poisoned environments, compared to state-of-the-art robust baselines.

Map generalization is commonly used in many GIS applications to produce maps with less detail so as to reduce the mapping complexity. Different from common simplifying strategies which simplify individual geometry objects separately, in this paper we consider the problem of generalizing the geometry objects under the topological constraints among the geometries and given constraining points. We propose a Cross-line algorithm to simplify the map while preserving the topological constraints. The proposed algorithm is extensively evaluated on five real map datasets and large synthetic datasets, and the results show that our proposed approach can greatly simplify the map with extremely high correctness and excellent efficiency.

Map Generalization is one of the most fundamental technologies for modern digital maps. It can effectively reduce the storage space and fit to different applications according to their scale requirement. This paper presents an efficient solution for this problem that won the ACM SIGSPATTAL CUP 2014. Given the original geometries which are represented by sampling points sequence, this method divides the boundaries into many small segments based on their topological characteristics and constriants. It attempts to minimize the number of sampling points by simplifying the given map and constraining points. In addition, the method also employs many optimization techniques to reduce the total latency, like memory pool, parallel computing and string parsing. Experimental results on real datasets demonstrate the effectiveness and efficiency of the proposed method.

The availability of massive trajectory data collected from GPS devices has received significant attentions in recent years. A hot topic is trip recommendation, which focuses on searching trajectories that connect (or are close to) a set of query locations, e.g., several sightseeing places specified by a traveller, from a collection of historic trajectories made by other travellers. However, if we know little about the sample coverage of trajectory data when developing an application of trip recommendation, it is difficult for us to answer many practical questions, such as 1) how many (future) queries can be supported with a given set of raw trajectories? 2) how many trajectories are required to achieve a good-enough result? 3) how frequent the update operations need to be performed on trajectory data to keep it long-term effective? In this paper, we focus on studying the overall quality of trajectory data from both spatial and temporal domains and evaluate proposed methods with a real big trajectory dataset. Our results should be useful for both the development of trip recommendation systems and the improvement of trajectory-searching algorithms.

In the age of big data, distributed computing frameworks tend to coexist and collaborate in pipeline using one scheduler. While a variety of techniques for reducing I/O latency have been proposed, these are rarely specific for the whole pipeline performance. This paper proposes memory management logic called “Hippo” which targets distributed systems and in particular “pipelined” applications that might span differing big data frameworks. Though individual frameworks may have internal memory management primitives, Hippo proposes to make a generic framework that works agnostic of these highlevel operations. To increase the hit ratio of in-memory cache, this paper discusses the granularity of caching and how Hippo leverages the job dependency graph to make memory retention and pre-fetching decisions. Our evaluations demonstrate that job dependency is essential to improve the cache performance and a global cache policy maker, in most cases, significantly outperforms explicit caching by users.

With the increasing availability of human-tracking data (e.g., Public transport IC card data, trajectory data, etc.), human mobility prediction is increasingly important. In this paper, we study a novel problem of using human-tracking data to predict human mobility and to detect over-crowded stations in public transport networks, and then finding unobstructed routes to go around these over-crowded stations. We believe that this study can bring significant benefits to users in many popular mobile applications such as route planning and recommendation, urban computing, and location based services in general. This problem is challenged by two difficulties: (1) how to detect crowded stations effectively, and (2) how to find unobstructed routes in public transport networks efficiently. To overcome these difficulties, we propose three human-mobility prediction methods based on uniform distribution, standard normal distribution, and priority ranking, respectively, to predict human mobility and to detect over-crowded stations. Then, we develop an efficient algorithm based on network expansion to find unobstructed routes in public transport networks. The performance of the developed algorithms has been verified by extensive experiments.

Trajectory search plays an important role in various applications such as trip planning and recommendation. However, most existing studies only focus on spatial proximity but ignore individual users' preferences. For example, it is inappropriate to recommend a route containing gravel roads to travelers without off-road vehicles. To accommodate various user preferences, we develop MOIR/UOTS, a trip recommendation system that supports User Oriented Trajectory Search (UOTS) [5], [7] based on our previous MOIR platform [2], [4]. Meanwhile, MOIR/UOTS also functions as a test-bed system for exploring and evaluating novel trajectory searching algorithms. In particular, we demonstrate how ordinary users can interact with MOIR/UOTS to search for trajectories with their preferences, and how MOIR/UOTS helps researchers to evaluate the performance of various algorithms.

The effectiveness of map-matching algorithms highly depends on the accuracy and correctness of underlying road networks. In practice, the storage capacity of certain hardware, e.g. mobile devices and embedded systems, is sometimes insufficient to maintain a large digital map for map-matching. Unfortunately, most existing map-matching approaches consider little about this problem. They only apply to environments with information-rich maps, but turn out to be unacceptable for map-matching on simplified road networks. In this paper, we propose a novel map-matching algorithm called Passby to work on most simplified road networks. The storage size of a digital map in disk or memory can be greatly reduced after the simplification. Even under the most simplified situation, i.e., each road segment only consists of a couple of intersection points and omits any other information of it, the experimental results on real dataset show that our Passby algorithm significantly maintains high matching accuracy. Benefiting from the small size of map, simple index structure and heuristic foresight strategy, Passby improves matching accuracy as well as efficiency.

Floating Car Data (FCD) provides an economic complement to infrastructure-based traffic monitoring systems. Based on our previous MOIR platform [5], we use FCD as the data source for large-scale real-time traffic monitoring. This new function brings a challenge of efficiently handling of streaming data from a very large number of moving objects. Server overload problems can occur when a system fails to process data and queries in real-tme, which can lead to critical issues such as unbounded delay accumulation, lost monitoring accuracy or lack of spontaneity. These problems can be addressed by adopting suitable load dropping decisions. In this work, we demonstrate several load shedding techniques, focusing on decision-making based on data attributes. With the end results being quantified and visualized using real data for a large city, this proof-of-concept system provides a convincing way of validating our ideas.

MOIR is a Web-based prototype to support a number of novel applications with network-constrained moving object management. Technical aspects of MOIR range from data acquisition and trajectory smoothing, trajectory data management and query processing, movement predications, and location based Web page recommendation. The demo is in the context of detailed digital road maps with 38,0000 road segments and 55,000 road intersection points, and real spatiotemporal data of over ten thousand taxis in Beijing.