Ziqiao Xi, Shuang Liang, Qi Liu, Jiaqing Zhang, Letian Peng, Fang Nan, Meshal Nayim, Tianhui Zhang, Rishika Mundada, Lianhui Qin, Biwei Huang, Kun Zhou. Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2026.

Conversational agents are required to respond to their users not only with high quality (i.e.commonsense-bearing) responses, but also considering multiple plausible alternative scenarios, reflecting the diversity in their responses.Despite the growing need to train diverse commonsense generators, the progress of this line of work has been significantly hindered by the lack of large-scale high-quality diverse commonsense training datasets.Due to the high annotation costs, existing Generative Commonsense Reasoning (GCR) datasets are created using a small number of human annotators, covering only a narrow set of commonsense scenarios.To address this training resource gap, we propose a two-stage method to create CommonSyn, the first-ever synthetic dataset for diversified GCR.Large Language Models (LLMs) fine-tuned on CommonSyn show simultaneous improvements in both generation diversity and quality compared with vanilla models and models fine-tuned on manually annotated datasets.1

Shamsuddeen Hassan Muhammad, Nedjma Ousidhoum, Idris Abdulmumin, Jan Philip Wahle, Terry Ruas, Meriem Beloucif, Christine de Kock, Nirmal Surange, Daniela Teodorescu, Ibrahim Said Ahmad, David Ifeoluwa Adelani, Alham Fikri Aji, Felermino D. M. A. Ali, Ilseyar Alimova, Vladimir Araujo, Nikolay Babakov, Naomi Baes, Ana-Maria Bucur, Andiswa Bukula, Guanqun Cao, Rodrigo Tufiño, Rendi Chevi, Chiamaka Ijeoma Chukwuneke, Alexandra Ciobotaru, Daryna Dementieva, Murja Sani Gadanya, Robert Geislinger, Bela Gipp, Oumaima Hourrane, Oana Ignat, Falalu Ibrahim Lawan, Rooweither Mabuya, Rahmad Mahendra, Vukosi Marivate, Alexander Panchenko, Andrew Piper, Charles Henrique Porto Ferreira, Vitaly Protasov, Samuel Rutunda, Manish Shrivastava, Aura Cristina Udrea, Lilian Diana Awuor Wanzare, Sophie Wu, Florian Valentin Wunderlich, Hanif Muhammad Zhafran, Tianhui Zhang, Yi Zhou, Saif M. Mohammad. Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2025.

In commonsense generation, given a set of input concepts, a model must generate a response that is not only commonsense bearing, but also capturing multiple diverse viewpoints. Numerous evaluation metrics based on form- and content-level overlap have been proposed in prior work for evaluating the diversity of a commonsense generation model. However, it remains unclear as to which metrics are best suited for evaluating the diversity in commonsense generation. To address this gap, we conduct a systematic meta-evaluation of diversity metrics for commonsense generation. We find that form-based diversity metrics tend to consistently overestimate the diversity in sentence sets, where even randomly generated sentences are assigned overly high diversity scores. We then use an Large Language Model (LLM) to create a novel dataset annotated for the diversity of sentences generated for a commonsense generation task, and use it to conduct a meta-evaluation of the existing diversity evaluation metrics. Our experimental results show that content-based diversity evaluation metrics consistently outperform the form-based counterparts, showing high correlations with the LLM-based ratings. We recommend that future work on commonsense generation should use content-based metrics for evaluating the diversity of their outputs.

As a promising technology for 6G networks, ultra-massive multiple-input multiple-output (UM-MIMO) faces substantial problems such as wavefront mismatch, high computational complexity, and reliance on unrealistic assumptions in hybrid field channel estimation. To address these issues, we propose a novel subarray-based measurement optimization framework. Initially, we adopt a wavefront decomposition model based on subarray. By decomposing the spherical wavefront into localized planar wavefront approximations, this model effectively resolves the wavefront mismatch and eliminates the dependence on the ratio factor while still ensuring high accuracy. Subsequently, a noise-robust measurement optimization algorithm is proposed to achieve a dual reduction in complexity. Specifically, a dynamic signal-to-noise ratio (SNR)-aware measurement optimization approach, utilizing heap sorting, prioritizes high-confidence measurements. Meanwhile, the two-dimensional inverse fast Fourier transform (2D-IFFT) estimation method reduces the unit computational load through parallel subarray processing. Simulation results demonstrate that our proposed algorithm consistently outperforms existing approaches across diverse SNR levels and ratio factors, with significant advantages in low-SNR scenarios. Additionally, in 2D array deployments, it shows an improvement of 3 dB in channel estimation error performance, highlighting its robustness and scalability in both configurations.

The smart distribution grid (SDG), characterized by large-scale interconnections and strong dependence on information and communication technologies, is highly susceptible to potential security threats, such as spoofing attacks and man-in-the-middle attacks. These threats may lead to the leakage of sensitive user power-expenditure information, even cause great economic damage. Therefore, authentication is of utmost importance in guaranteeing the electrical safety of SDGs. In this paper, we present a distributed physical layer authentication (DPLA) scheme tailored for smart meter authentication. The scheme overcomes the limitations of traditional upper-layer cryptography-based mechanisms, and achieves lightweight continuous authentication in a cooperative manner. To fully exploit the channel information collected by collaborative nodes located in different azimuths, a CNN algorithm is designed for deep feature extraction. Moreover, a situational-aware dynamic weighted voting strategy is introduced to coordinate inconsistent opinions, thereby making unified decisions. Aimed at maximizing the integrated performance gains of DPLA, both long-term reputation and short-term performance are taken into account for node’s weight update. Finally, simulations are carried out. The results demonstrate that our scheme outperforms DPLAs based on static voting strategies with respect to authentication accuracy, anti-disturbance robustness and environmental adaptability; Hence, it caters to the demand for high-quality continuous authentication in SDGs.

Machine learning (ML)-based physical layer authentication (PLA) has attracted much attention since neural networks can be constructed to identify channel characteristics in complex wireless environments. This enables high-authentication performance and lightweight deployment in the Internet of Things (IoTs). Due to the booming growth of IoT connections, the workload of the central authenticator increases significantly. As a result, resource-constrained terminals are unable to independently handle the computationally complex ML task. Therefore, cooperative PLA (CoPLA), which introduces multiple supervised nodes as task-sharing entities, is emerged as a promising solution to address this concern. However, in existing CoPLA studies, some critical issues have been overlooked. For example, the consideration of which collaborative nodes are eligible or best suited for cooperation to maximize the authentication gains. Moreover, the security threats posed by untrusted collaborators are equally challenging. In this article, we propose a federated learning (FL)-based CoPLA scheme that utilizes a group of edge devices to jointly build an authenticator. This ensures privacy preservation and higher robustness. To figure out the optimal collaborator selection in CoPLA, an adaptive search procedure via reinforcement learning (RL) is customized. Furthermore, we introduce a lightweight reputation estimation method to evaluate each collaborator’s credibility, thereby uncovering underperforming devices or hidden internal attackers. Finally, simulations and real-world experiments are carried out. The results show that the authentication accuracy of our scheme is 9.52% higher than that of blind cooperation. And, it outperforms other existing CoPLA schemes in terms of time efficiency and robustness.