Complex games have long been an important benchmark for testing the progress of artificial intelligence algorithms. AlphaGo, AlphaZero, and MuZero have defeated top human players in Go and Chess, garnering widespread societal attention towards artificial intelligence. Concurrently, large language models (LLMs) have exhibited remarkable capabilities across various tasks, raising the question of whether LLMs can achieve similar success in complex games. In this paper, we explore the potential of LLMs in mastering complex card games. We systematically assess the learning capabilities of LLMs across eight diverse card games, evaluating the impact of fine-tuning on high-quality gameplay data, and examining the models' ability to retain general capabilities while mastering these games. Our findings indicate that: (1) LLMs can approach the performance of strong game AIs through supervised fine-tuning on high-quality data, (2) LLMs can achieve a certain level of proficiency in multiple complex card games simultaneously, with performance augmentation for games with similar rules and conflicts for dissimilar ones, and (3) LLMs experience a decline in general capabilities when mastering complex games, but this decline can be mitigated by integrating a certain amount of general instruction data. The evaluation results demonstrate strong learning ability and versatility of LLMs. The code is available at https://github.com/THUDM/LLM4CardGame

MOTIVATION: As one of the recalcitrant challenges in life sciences and biomedicine, protein function prediction suffers from a deluge of AI-designed proteins, particularly having to face multi-modal information in the era of big data. Importing the high-throughput neural-network-based prediction framework to replace the low-throughput biological experiments, a universal multi-modal method is straightforward in addressing the growing gap between known sequences and predicting functions. RESULTS: To bridge the gap, we propose ProtGO, a three-step framework for predicting protein function, which leverages the credible Gene Ontology (GO) knowledge base and integrates four common modalities. Specifically, we first introduce frontier pre-trained protein language models (PLMs) for representation learning of mainstay functional protein sequences. For the remaining multi-modal data, we design a text alignment module for explainable text descriptions, a taxonomy encoding module for species-specific taxonomy, and a GO graph embedding module for biological GO relations. Each module is independent and adaptive for the referenced modalities. By harnessing these four knowledge representations, ProtGO maximizes the potential of GO resources, enhancing the performance of vanilla PLMs and biological language models (LMs) in downstream GO prediction tasks. Extensive experiments demonstrate that ProtGO significantly advances the abilities of state-of-the-art PLMs to predict protein functions: approximately 8% to 27% increase in the maximum F1 measure (Fmax) compared to base models. These comprehensive studies confirm ProtGO's capability to deliver outstanding performance in protein function prediction by utilizing a rich blend of functional and evolutionary knowledge. AVAILABILITY AND IMPLEMENTATION: Our source code and all the data are available at https://github.com/sunyatawang/ProtGO.

Abstract The forecasting of steel futures prices is important for the steel futures market, even for the steel industry. We propose a decomposition ensemble model that incorporates the Ensemble Empirical Mode Decomposition (EEMD), Long Short‐Term Memory (LSTM), Support Vector Regression (SVR), and Back Propagation (BP) neural network to forecast steel futures prices. The forecasting procedures are as follows: (1) The price data are initially decomposed into several relatively independent Intrinsic Mode Functions (IMFs) and a residue using EEMD. (2) The IMFs are then reconstructed as components representing short‐term, medium‐term, and long‐term frequencies via fine‐to‐coarse. (3) LSTM, SVR, and BP neural network are utilized to forecast the short‐term, medium‐term, and long‐term reconstructed components, respectively. (4) The prediction results for each component are simply added to the final prediction results. The accuracy of the proposed model is compared with several benchmark models by experiments and evaluated by some prediction evaluation indexes. The experimental results show that our model outperforms other models in terms of forecast accuracy, confirming its strong predictive capabilities. This study provides some suggestions for investment and decision making by participants in the steel futures market. It may promote the smooth operation of the steel futures market and shed some light on the operation of the steel industry.

Group buying is a business model where people with the same merchandise interests form a group and conduct the purchase together to achieve a discount. Third-party proxy websites negotiate with merchants for appealing deals and then provide them to end customers. We call it online group buying. Besides, there exists local group buying where the joiners, the initiator, and sometimes even the merchants are in the same local community. Such locality induces some interesting characteristics in group buying, which remain largely unexplored in the research community. This study attempts to reveal users' behaviors in group buying within the local context. We developed a mobile service called "HappyGo" that supports local group buying. We conducted a trial involving more than 300 users from a company office. From our findings, we believe that local group buying complements online group buying by creating a "local" economic circle while also providing users with social benefits.

We study the integrated WiFi/WiMAX networks, where users are equipped with dual-radio interfaces that can connect to either a WiFi or a WiMAX network. Previous research on integrated heterogeneous networks (e.g., WiFi/cellular) usually considers one network as the main and the other as the auxiliary. The performance of the integrated network is compared with the "main” network. The gain is apparently due to the additional resources from the auxiliary network. In this study, we are interested in integration gain that comes from the better utilization of the resource rather than the increase of the resource. The heterogeneity of the two networks is the fundamental reason for the integration gain. To quantify it, we design a generic framework that supports different performance objectives. We focus on the max-min throughput fairness in this work and also briefly cover the proportional fairness metric. We first prove that it is NP-hard to achieve integral max-min throughput fairness, then propose a heuristic algorithm, which provides two-approximation to the optimal fractional solution. Simulation results demonstrate significant integration gain from three sources, namely, spatial multiplexing, multinetwork diversity, and multiuser diversity. For the proportional fairness metric, we derive the formulation and propose a heuristic algorithm, which shows satisfactory performance when compared with the optimal solution.

Joint routing-and-scheduling has been considered in wireless mesh networks for its significant performance improvement. While existing work assumes it, accurate traffic information is usually not available due to traffic dynamics, as well as inaccuracy and delay in its measurement and dissemination. In addition, the joint routing and scheduling usually requires a centralized controller to calculate the optimal routing and scheduling and distribute such policies to all the nodes. Thus, even if the accurate traffic information is always available, the central controller has to compute the routing and scheduling repeatedly because the traffic demands change continuously. This leads to prohibitive computation and distribution overhead. Therefore, in this paper, we propose a joint routing-scheduling scheme that achieves robust performance under traffic information uncertainty. In particular, it achieves worst-case optimal performance under a range of traffic conditions. This unique feature validates the use of centralized routing and scheduling in wireless mesh networks. As long as the traffic variation is within the estimation range, the routing and scheduling do not need to be recomputed and redistributed. Through extensive simulations, we show that our proposed scheme meets the objective (i.e., optimizes the worst-case performance). Moreover, although it only guarantees the worst-case performance in theory, its average performance is also good. For example, our proposed scheme can perform better than a fixed optimal routing and scheduling scheme in more than 80 percent of 500 random traffic instances. Our scheme provides insights on the desired properties of multipath routing, namely, spatial reuse and load balancing.

Abstract — Wireless networks with multi-channel multi-radio availability are attracting more and more attention from the research community because of its performance improvement and relatively low cost and complexity. Most work in the literature consider the performance of multi-channel multiradio wireless networks with predefined numbers of channels and radios, and develop algorithms for channel allocation and transmission scheduling. The capacity limit on multi-channel multi-radio wireless networks was seldom addressed before. In this work, we study two problems: In a specific topology, 1) what is the maximum capacity we can get given the number of channels and radios? and 2) what is the impact of the number of radios on the system performance? To answer the above questions, we propose a general framework to find the maximum capacity given a multi-channel multi-radio wireless network. Its result also provides an indication of the “goodness ” of a topology. We then use the framework to study the impact of radio constraints. I.

The lack of end-to-end quality of service support in the current Internet has caused significant difficulties to ensuring playback continuity in video streaming applications. This study addresses this challenge by investigating a new adaptation algorithm to adjust the bit-rate of video data in response to the network bandwidth available to improve playback continuity. Unlike previous works, the proposed algorithm is transparent to the video client, requires no parameter tuning, and yet can outperform existing algorithms. This paper presents this algorithm, evaluates and compares its performance with the best algorithm currently available using extensive trace-driven simulations.

Voice-over-IP (VoIP) is.an important application on the Internet. With the emergence of WLAN technology and its various advantages compared with the traditional wired LAN, it is fast becoming the "last-mile" of choice for the overall Internet infrastructure. This work considers the support of VoIP over 802.11b WLAN. We show that although the raw WLAN capacity can potentially support more than 500 VoIP sessions, various overheads bring this down to only 12 VoIP sessions when using GSM 6.10 codec. We propose a novel multiplexing scheme for VoIP which exploits multicasting over WLAN for the downlink VoIP traffic. This scheme can achieve nearly 100% improvement in system capacity. In addition, we present results showing that the delay and delay jitter introduced by the proposed scheme are small. We believe that the scheme can reduce the blocking probability of VoIP sessions in an enterprise WLAN significantly.

The rapid growth of wireless LAN (WLAN) deployments will bring about many novel mobile applications. Among them will be real-time multimedia streaming applications running on UDP, which may interfere with current data applications running on TCP. This paper is a first attempt to investigate how to ensure the performance of these two groups of applications when they co-exist over a WLAN. Toward this end, we have designed and implemented a UDP rate adaptation scheme called adaptive-buffer rate control (ABRC) for multimedia streaming over WLAN. ABRC has two distinguishing features compared with other schemes: 1) it can achieve arbitrary bandwidth allocations between UDP and TCP in the WLAN, as opposed Io previously proposed "TCP friendly" schemes, which can only achieve uniform bandwidth allocations; and 2) the majority of previously proposed flexible bandwidth-allocation schemes achieve arbitrary bandwidth allocations by prioritizing and scheduling packet transmissions within network equipment (i.e., within routers, base stations, etc.). In contrast, ABRC is an end-to-end application-layer solution that does not require changes to current WLAN products, making it more readily deployable over existing networks.

This paper examines the impact of RED on two versions of TCP - traditional TCP Reno and a newly proposed variant, TCP Veno - over 802.11b WLAN. TCP Reno was originally designed for wired networks where packet losses are primarily due to network congestion. This assumption is not always true in wireless networks, in which packet losses can be due to transmission errors on the noisy wireless link. TCP Veno refines the algorithms in Reno by distinguishing between noncongestive and congestive states, and avoids the unnecessary reduction of TCP congestion window when packet losses are not due to congestion. Our results show that TCP Veno can achieve up to 30% more throughput than TCP Reno when link quality is poor. Our results also show that TCP Veno is compatible with RED. In addition, although RED does not help to further improve the throughput in Veno, it can improve fairness among co-existing TCP flows.