Hai Wan, Haicheng Chen, Jianfeng Du, Weilin Luo, Rongzhen Ye. Proceedings of the 59th Annual Meeting of the Association for Computational Linguistics and the 11th International Joint Conference on Natural Language Processing (Volume 1: Long Papers). 2021.

Recommender systems are widely used to help users locate information by suggesting goods that best match users’ needs and preferences. Most existing recommender systems are built by implementing supervised learning-based algorithms. Unfortunately, the large search space with lots of local optimums makes it challenging for the existing recommender systems to apply in more and more new upcoming recommendation scenarios, such as the novel item-list recommendation scenario proposed by FUXI AI Lab, Netease. It requires to recommend an item list in each step, where the item list is one of the thousands of combinations of items. To tackle these challenges, recent researchers resort to adopting reinforcement learning (RL) for recommendations. It is natural to formalize item recommendation problem as a multi-step decision-making problem and use exploration and exploitation of RL to address the problem of large search space. In this paper, we propose a DQN-based recommender system for item-list recommendation that achieves second place in track II of the RL-based RecSys and it only performs slightly worse than the first place.1

Control firmware in unmanned aerial vehicles (UAVs) uses sensors to model and manage flight operations, from takeoff to landing to flying between waypoints. However, sensors can fail at any time during a flight. If control firmware mishandles sensor failures, UAVs can crash, fly away, or suffer other unsafe conditions. In-situ model checking finds sensor failures that could lead to unsafe conditions by systematically failing sensors. However, the type of sensor failure and its timing within a flight affect its manifestation, creating a large search space. We propose Avis, an in-situ model checker to quickly uncover UAV sensor failures that lead to unsafe conditions. Avis exploits operating modes, i.e., a label that maps software execution to corresponding flight operations. Widely used control firmware already support operating modes. Avis injects sensor failures as the control firmware transitions between modes - a key execution point where mishandled software exceptions can trigger unsafe conditions. We implemented Avis and applied it to ArduPilot and PX4. Avis found unsafe conditions 2.4X faster than Bayesian Fault Injection, the leading, state-of-theart approach. Within the current code base of ArduPilot and PX4, Avis discovered 10 previously unknown software bugs that lead to unsafe conditions. Additionally, we reinserted 5 known bugs that caused serious, unsafe conditions and Avis correctly reported all of them.

The Maximum k-plex Problem is an important combinatorial optimization problem with increasingly wide applications. In this paper, we propose a novel strategy, named Dynamic-threshold Configuration Checking (DCC), to reduce the cycling problem of local search. Due to the complicated neighborhood relations, all the previous local search algorithms for this problem spend a large amount of time in identifying feasible neighbors in each step. To further improve the performance on dense and challenging instances, we propose Double-attributes Incremental Neighborhood Updating (DINU) scheme which reduces the worst-case time complexity per iteration from O(|V|⋅ΔG) to O(k · Δ‾G). Based on DCC strategy and DINU scheme, we develop a local search algorithm named DCCplex. According to the experiment result, DCCplex shows promising result on DIMACS and BHOSLIB benchmark as well as real-world massive graphs. Especially, DCCplex updates the lower bound of the maximum k-plex for most dense and challenging instances.

Network partitions are inevitable in large-scale cloud systems. Despite developer's efforts in handling network partitions throughout designing, implementing and testing cloud systems, bugs caused by network partitions, i.e., partition bugs, still exist and cause severe failures in production clusters. It is challenging to expose these partition bugs because they often require network partitions to start and stop at specific timings.

Exception mechanism is widely used in cloud systems. This is mainly because it separates the error handling code from main business logic. However, the huge space of potential error conditions and the sophisticated logic of cloud systems present a big hurdle to the correct use of exception mechanism. As a result, mistakes in the exception use may lead to severe consequences, such as system downtime and data loss. To address this issue, the communities direly need a better understanding of the exception-related bugs, i.e., eBugs, which are caused by the incorrect use of exception mechanism, in cloud systems. In this paper, we present a comprehensive study on 210 eBugs from six widely-deployed cloud systems, including Cassandra, HBase, HDFS, Hadoop MapReduce, YARN, and ZooKeeper. For all the studied eBugs, we analyze their triggering conditions, root causes, bug impacts, and their relations. To the best of our knowledge, this is the first study on eBugs in cloud systems, and the first one that focuses on triggering conditions. We find that eBugs are severe in cloud systems: 74% of our studied eBugs affect system availability or integrity. Luckily, exposing eBugs through testing is possible: 54% of the eBugs are triggered by non-semantic conditions, such as network errors; 40% of the eBugs can be triggered by simulating the triggering conditions at simple system states. Furthermore, we find that the triggering conditions are useful for detecting eBugs. Based on such relevant findings, we build a static analysis tool, called DIET, and apply it to the latest versions of the studied systems. Our results show that DIET reports 31 bugs and bad practices, and 23 of them are confirmed by the developers as "previously-unknown" ones.

Software should behave correctly even in adverse conditions. Particularly, we study the problem of automated validation of crash consistency, i.e., file system data safety when systems crash. Existing work requires non-trivial manual efforts of specifying checking scripts and workloads, which is an obstacle for software developers. Therefore, we propose C3, a novel approach that makes crash consistency validation as easy as pressing a single button. With a program and an input, C3 automatically reports inconsistent crash sites. C3 not only exempts developers from the need of writing crash site checking scripts (by an algorithm that computes editing distance between file system snapshots) but also reduces the reliance on dedicated workloads (by test amplification). We implemented C3 as an open-source tool. With C3, we found 14 bugs in open-source software that have severe consequences at crash and 11 of them were previously unknown to the developers, including in highly mature software (e.g., GNU zip and GNU coreutils sort) and popular ones being actively developed (e.g., Adobe Brackets and TeXstudio).