Counterfactual reasoning is widely recognized as one of the most challenging and intricate aspects of causality in artificial intelligence. In this paper, we evaluate the performance of large language models (LLMs) in counterfactual reasoning. In contrast to previous studies that primarily focus on commonsense causal reasoning, where LLMs often rely on prior knowledge for inference, we specifically assess their ability to perform counterfactual inference using a set of formal rules. To support this evaluation, we introduce a new benchmark dataset, CounterBench, comprising 1.2K counterfactual reasoning questions. The dataset is designed with varying levels of difficulty, diverse causal graph structures, distinct types of counterfactual questions, and multiple nonsensical name variants. Our experiments demonstrate that counterfactual reasoning poses a significant challenge for LLMs, with most models performing at levels comparable to random guessing. To enhance LLM's counterfactual reasoning ability, we propose a novel reasoning paradigm, CoIn, which guides LLMs through iterative reasoning and backtracking to systematically explore counterfactual solutions. Experimental results show that our method significantly improves LLM performance on counterfactual reasoning tasks and consistently enhances performance across different LLMs.

Charging processes are the key to promoting electric taxis and improving their operational efficiency due to frequent charging activities and long charging time. Nevertheless, optimizing charging resource allocation in real time is extremely challenging because of uneven charging demand/supply distributions, heuristic-based charging behaviors of drivers, and city-scale of the fleets. The existing solutions have utilized real-time contextual information for charging recommendation, but they do not consider the much-richer fleet information, leading to the suboptimal individual-based charging recommendation. In this paper, we design a data-driven fleet-oriented charging recommendation system for charging resource allocation called ForETaxi for electric taxis , which aims to minimize the overall charging overhead for the entire fleet, instead of individual vehicles. ForETaxi considers not only current charging requests but also possible charging requests of other nearby electric taxis in the near future by inferring their status in real time. More importantly, we implement ForETaxi with multiple types of sensor data from the Chinese Shenzhen city including GPS data, and taxi transaction data from more than 13,000 electric taxis, combined with road network data and charging station data. The data-driven evaluation results show that compared to the state-of-the-art individual-based recommendation methods, our fleet-oriented ForETaxi outperforms them by 16% in the total charging time reduction and 82% in the queuing time reduction.