arXiv subjects

Shijie Dai

Publications and source records attributed to Shijie Dai.

2 recordsLinked to original sources

arXiv2026

Evaluating Investment Logic in Large Language Models: A Real-World Benchmark Towards Personalzied Financial Agents

Investment competence is inherently personalized: the same market evidence can justify different actions for investors with different goals, horizons, portfolios, and risk boundaries. Yet financial LLMs are evaluated either by static question answering or by terminal profit and loss. The former omits agency; the latter cannot reveal whether a profitable action was grounded, profile-consistent, or merely lucky. We ask whether the community is using the wrong ruler for consequential agents. We introduce \textsc{InvestLogicBench}, a process-native benchmark containing 201,247 documented decisions from 151 real-world investors. Each episode instantiates a \textbf{P$\rightarrow$E$\rightarrow$R$\rightarrow$D$\rightarrow$O} trace: investor \textit{Profile}, observable market \textit{Events}, investment \textit{Reasoning}, executable \textit{Decision}, and delayed \textit{Outcome}. The release includes profile construction, point-in-time event binding, structured logic, horizons, outcomes, and post-mortems, and supports comprehension, profile-conditioned generation, and end-to-end replay. Across four leading LLMs, logical plausibility remains near 4/5 while event grounding is only 0.8--2.8/5; return and process quality also disagree. These results expose polished but weakly grounded reasoning that outcome-only evaluation hides. We further argue that P$\rightarrow$E$\rightarrow$R$\rightarrow$D$\rightarrow$O should be a data-system interface, requiring versioned profiles, temporal provenance, inspectable retrieval, decision ledgers, and replayable outcomes. Finance is our stress test for a broader class of personalized, consequential agents.

cs.AI↗

arXiv2018

A Computation Offloading Incentive Mechanism with Delay and Cost Constraints under 5G Satellite-ground IoV architecture

The 5G Internet of Vehicles has become a new paradigm alongside the growing popularity and variety of computation-intensive applications with high requirements for computational resources and analysis capabilities. Existing network architectures and resource management mechanisms may not sufficiently guarantee satisfactory Quality of Experience and network efficiency, mainly suffering from coverage limitation of Road Side Units, insufficient resources, and unsatisfactory computational capabilities of onboard equipment, frequently changing network topology, and ineffective resource management schemes. To meet the demands of such applications, in this article, we first propose a novel architecture by integrating the satellite network with 5G cloud-enabled Internet of Vehicles to efficiently support seamless coverage and global resource management. A incentive mechanism based joint optimization problem of opportunistic computation offloading under delay and cost constraints is established under the aforementioned framework, in which a vehicular user can either significantly reduce the application completion time by offloading workloads to several nearby vehicles through opportunistic vehicle-to-vehicle channels while effectively controlling the cost or protect its own profit by providing compensated computing service. As the optimization problem is non-convex and NP-hard, simulated annealing based on the Markov Chain Monte Carlo as well as the metropolis algorithm is applied to solve the optimization problem, which can efficaciously obtain both high-quality and cost-effective approximations of global optimal solutions. The effectiveness of the proposed mechanism is corroborated through simulation results.

cs.NI↗