Sihong Wu, Owen Jiang, Yilun Zhao, Tiansheng Hu, Yiling Ma, Kaiyan Zhang, Manasi Patwardhan, Arman Cohan. Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2026.

Tianyu Yang, Sihong Wu, Yilun Zhao, Zhenwen Liang, Lisen Dai, Chen Zhao, Minhao Cheng, Arman Cohan, Xiangliang Zhang. Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2026.

Enabling models to recognize vast open-world categories has been a longstanding pursuit in object detection. By leveraging the generalization capabilities of vision-language models, current open-world detectors can recognize a broader range of vocabularies, despite being trained on limited categories. However, when the scale of the category vocabularies during training expands to a real-world level, previous classifiers aligned with coarse class names significantly reduce the recognition performance of these detectors. In this paper, we introduce Prova, a multi-modal prototype classifier for vast-vocabulary object detection. Prova extracts comprehensive multi-modal prototypes as initialization of alignment classifiers to tackle the vast-vocabulary object recognition failure problem. On V3Det, this simple method greatly enhances the performance among one-stage, two-stage, and DETR-based detectors with only additional projection layers in both supervised and open-vocabulary settings. In particular, Prova improves Faster R-CNN, FCOS, and DINO by 3.3, 6.2, and 2.9 AP respectively in the supervised setting of V3Det. For the open-vocabulary setting, Prova achieves a new state-of-the-art performance with 32.8 base AP and 11.0 novel AP, which is of 2.6 and 4.3 gain over the previous methods.

We present SciArena, an open and collaborative platform for evaluating foundation models on scientific literature-grounded tasks. Unlike traditional benchmarks for scientific literature understanding and synthesis, SciArena engages the research community directly, following the Chatbot Arena evaluation approach of community voting on model comparisons.By leveraging collective intelligence, SciArena offers a community-driven evaluation of model performance on open-ended scientific tasks that demand literature-grounded, long-form responses.The platform currently supports 44 open-source and proprietary foundation models and has collected over 19,000 votes from human researchers across diverse scientific domains. Our analysis of the data collected so far confirms its high quality.We discuss the results and insights based on the model ranking leaderboard.To further promote research in building model-based automated evaluation systems for literature tasks, we release SciArena-Eval, a meta-evaluation benchmark based on our collected preference data. The benchmark measures the accuracy of models in judging answer quality by comparing their pairwise assessments with human votes. Our experiments highlight the benchmark’s challenges and emphasize the need for more reliable automated evaluation methods.

The ultra-low-frequency (ULF) electromagnetic anomaly has been considered one of the earthquake precursory signals with the potential for short-term prediction. As such, effectively detecting ULF anomalies is important for mitigating earthquake disasters. Given the comprehensive coverage of geomagnetic networks in Japan, the 2011 Tohoku-Oki earthquake (M9.0) provided a tremendous opportunity for investigating the characteristics and mechanisms of ULF anomalies. Previous studies reporting detections of ULF anomalies triggered by the Tohoku-Oki earthquake have been questioned on the grounds of insufficient reliability due to technological limitations. In this study, we employ a multi-reference station data-quality-weighted method to detect the ULF anomaly. Comparison with traditional single-reference station method demonstrates the robustness of our detection results. Statistical test also indicates that the ULF anomaly appearing approximately two months before the earthquake was driven by physical processes. Geomagnetic storm analysis further rules solar activity as the cause of the anomaly. Spatial distribution of the anomaly amplitude reveals a decrease in the anomaly energy with increasing epicentral distance, implying a strong association between the anomaly and the earthquake. Our study is helpful for understanding the possible connection between the ULF electromagnetic precursors and the seismogenic processes of strong earthquakes, which contributes to achieving the ultimate goal of short-term earthquake prediction.

The inversion of airborne electromagnetic (AEM) data suffers from severe non-uniqueness of the solution. Bayesian inference provides the means to estimate structural uncertainty with a rich suite of statistical information. However, conventional Bayesian methods are computationally demanding in nonlinear inversions, especially considering the huge volumes of observational data, and thus are not feasible in practice. In this study, we develop a fast Bayesian inversion operator based on the invertible neural network (INN) to fully explore the posterior distribution and quantitatively evaluate the model uncertainty. The INN uses a latent variable to capture the information loss during measurement and constructs bijective mappings between AEM data and the resistivity model. We also introduce another noise variable into the INN to account for data uncertainties. Synthetic tests demonstrate that the INN can effectively recover the posterior distribution by a relatively small ensemble of predicted resistivity models whose AEM responses show a significant agreement with the true signal. We also apply the INN inversion operator to a field data set and obtain results consistent with previous studies. The INN shows considerable adaptability to field observations and strong noise robustness. Meanwhile, the INN delivers the inversion result with posterior model distribution for 23366 AEM time series in 20 seconds on a common PC. The inversion efficiency can be further improved for large data set due to its natural parallelizability. The proposed INN method can support fast Bayesian inversion of AEM data and offer tremendous potential for near real-time uncertainty evaluation of underground structures.

The geoelectric data contain important anomalous information for short-term earthquake prediction. Timely and accurate identification of seismic electric anomalies is important for disaster prevention. However, identifying anomalies is challenging due to the huge volumes of data and noise disturbance. In this study, we develop a real-time automatic search engine (RASE) that incorporates an unsupervised convolutional denoising network (UCN) module and a supervised LSTM network (SLN) prediction module to automatically search for important anomalous signals in real-time. Experiments demonstrate that the RASE provides excellent detection accuracy and efficiency for synthetic and field data, which takes only dozens of seconds for a common PC to provide accurate detection results for data collected over a 24-hour period. The RASE has excellent flexibility and developability, as its internal modules can be adapted by more suitable technologies for better performance in various application scenarios. Comparison of multiple module combinations shows that the RASE configured with UCN and SLN has the highest detection accuracy. Our proposed search engine can reduce the human labor required for complex and repetitive detection work and fully realize the potential of geoelectric field observation in earthquake monitoring and disaster prevention.

Seismic electric signals (SESs) are essential short-term precursors of earthquakes. Accurate and efficient detection of SESs is significant to short-term predictions of earthquakes. However, SESs are usually disturbed by various noises and are thus difficult to recognize. Although conventional techniques have made substantial efforts in improving the SES detection accuracy, the success rates of SES detection at certain stations are still less satisfactory due to the complexity and diversity of noises. In this study, we apply deep learning to extract SESs and develop a novel deep learning network based on geoelectric field characteristics by combining the long short-term memory (LSTM) blocks with an autoencoder structure and a time-step attention module. The detection results of both synthetic and real data demonstrate that our proposed network yields superior performance in detecting embedded SESs in the presence of severe noise interference compared with traditional methods and several well-known networks. Moreover, our novel network shows the excellent ability of massive data processing, generalization and migration, which can process one-day’s worth of data in only milliseconds, adapt to SESs whose durations and amplitudes are different from those of the training set and be easily transferred to newly acquired data. The proposed novel method can provide more efficient and accurate detection results, which will broaden the data availability of hazard mitigation based on SESs.