LIMU-BERT
The DOI matches its Crossref record; 56 of the 56 references it cites were found in a registry or an index.
Abstract
Deep learning greatly empowers Inertial Measurement Unit (IMU) sensors for various mobile sensing applications, including human activity recognition, human-computer interaction, localization and tracking, and many more. Most existing works require substantial amounts of well-curated labeled data to train IMU-based sensing models, which incurs high annotation and training costs. Compared with labeled data, unlabeled IMU data are abundant and easily accessible. In this work, we present LIMU-BERT, a novel representation learning model that can make use of unlabeled IMU data and extract generalized rather than task-specific features. LIMU-BERT adopts the principle of self-supervised training of the natural language model BERT to effectively capture temporal relations and feature distributions in IMU sensor measurements. However, the original BERT is not adaptive to mobile IMU data. By meticulously observing the characteristics of IMU sensors, we propose a series of techniques and accordingly adapt LIMU-BERT to IMU sensing tasks. The designed models are lightweight and easily deployable on mobile devices. With the representations learned via LIMU-BERT, task-specific models trained with limited labeled samples can achieve superior performances. We extensively evaluate LIMU-BERT with four open datasets. The results show that the LIMU-BERT enhanced models significantly outperform existing approaches in two typical IMU sensing applications.
Overview
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
Claims other papers make about this one
Each is checked against this paper’s own text: does it say what the citing paper says it does?
In the citing paper
“On Opportunity and Realdisp datasets, LIMU-BERT [71] achieves the best performance.”
From Spatial-Temporal Masked Autoencoder for Multi-Device Wearable Human Activity Recognition (2023) · cites this paper as [71]
Citations
Cites
Hochreiter, S., & Schmidhuber, J. (1997) · Neural Computation · doi:10.1162/neco.1997.9.8.1735 · Cited by 102,318
BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding
Devlin, J., Chang, M.-W., Lee, K., & Toutanova, K. (2018) · arXiv · doi:10.48550/arxiv.1810.04805 · Cited by 73,173
(2008) · Journal of Machine Learning Research · Cited by 33,051
A survey on Image Data Augmentation for Deep Learning
Shorten, C., & Khoshgoftaar, T. M. (2019) · Journal of Big Data · doi:10.1186/s40537-019-0197-0 · Cited by 13,203
(2017) · Cited by 11,625
Empirical Evaluation of Gated Recurrent Neural Networks on Sequence Modeling
Chung, J., Gulcehre, C., Cho, K., & Bengio, Y. (2014) · arXiv · doi:10.48550/arxiv.1412.3555 · Cited by 11,008
Extracting and composing robust features with denoising autoencoders
Vincent, P., Larochelle, H., Bengio, Y., & Manzagol, P.-A. (2008) · Proceedings of the 25th international conference on Machine learning - ICML '08 · doi:10.1145/1390156.1390294 · Cited by 7,436
ALBERT: A Lite BERT for Self-supervised Learning of Language Representations
Lan, Z., Chen, M., Goodman, S., Gimpel, K., Sharma, P., & Soricut, R. (2019) · arXiv · doi:10.48550/arxiv.1909.11942 · Cited by 4,308 · cited 3×
“Second, LIMU-BERT adopts a cross-layer parameter sharing mechanism [19] to improve parameter efficiency.”
BackgroundMethodFederated Learning: Strategies for Improving Communication Efficiency
Konečný, J., McMahan, H. B., Yu, F. X., Richtárik, P., Suresh, A. T., & Bacon, D. (2016) · arXiv · doi:10.48550/arxiv.1610.05492 · Cited by 3,766
Gaussian Error Linear Units (GELUs)
Hendrycks, D., & Gimpel, K. (2016) · arXiv · doi:10.48550/arxiv.1606.08415 · Cited by 3,537
Research works printed but not linked: 6
- Jimmy Lei Ba , Jamie Ryan Kiros, and Geoffrey E. Hinton . 2016 . Layer Normalization . arXiv:1607.06450 [stat.ML] Jimmy Lei Ba, Jamie Ryan Kiros, and Geoffrey E. Hinton. 2016. Layer Normalization. arXiv:1607.06450 [stat.ML] · arXiv
- Diederik P Kingma and Jimmy Ba . 2014 . Adam: A method for stochastic optimization. arXiv preprint arXiv:1412.6980 (2014). Diederik P Kingma and Jimmy Ba. 2014. Adam: A method for stochastic optimization. arXiv preprint arXiv:1412.6980 (2014). · arXiv
- Alex Krizhevsky , Ilya Sutskever , and Geoffrey E Hinton . 2012. Imagenet classification with deep convolutional neural networks. Advances in neural information processing systems 25 ( 2012 ), 1097--1105. Alex Krizhevsky, Ilya Sutskever, and Geoffrey E Hinton. 2012. Imagenet classification with deep convolutional neural networks. Advances in neural information processing systems 25 (2012), 1097--1105.
- Brendan McMahan Eider Moore Daniel Ramage Seth Hampson and Blaise Aguera y Arcas. 2017. Communication-efficient learning of deep networks from decentralized data. In Artificial intelligence and statistics. PMLR 1273--1282. Brendan McMahan Eider Moore Daniel Ramage Seth Hampson and Blaise Aguera y Arcas. 2017. Communication-efficient learning of deep networks from decentralized data. In Artificial intelligence and statistics. PMLR 1273--1282.
- Ashish Vaswani , Noam Shazeer , Niki Parmar , Jakob Uszkoreit , Llion Jones , Aidan N Gomez , Lukasz Kaiser , and Illia Polosukhin . 2017. Attention is all you need. arXiv preprint arXiv:1706.03762 ( 2017 ). Ashish Vaswani, Noam Shazeer, Niki Parmar, Jakob Uszkoreit, Llion Jones, Aidan N Gomez, Lukasz Kaiser, and Illia Polosukhin. 2017. Attention is all you need. arXiv preprint arXiv:1706.03762 (2017). · arXiv
- Jianbo Yang , Minh Nhut Nguyen , Phyo Phyo San , Xiaoli Li , and Shonali Krishnaswamy . 2015 . Deep convolutional neural networks on multichannel time series for human activity recognition .. In Ijcai , Vol. 15 . Buenos Aires, Argentina, 3995--4001. Jianbo Yang, Minh Nhut Nguyen, Phyo Phyo San, Xiaoli Li, and Shonali Krishnaswamy. 2015. Deep convolutional neural networks on multichannel time series for human activity recognition.. In Ijcai, Vol. 15. Buenos Aires, Argentina, 3995--4001.
Cited by
Items 101–144 of 144
SelfReplay: Adapting Self-Supervised Sensory Models via Adaptive Meta-Task Replay
Yoon, H., Kwak, J., Tolera, B. A., Dai, G., Li, M., Gong, T., Lee, K., & Lee, S.-J. (2025) · ACM Conference on Embedded Networked Sensor Systems (SenSys) · doi:10.1145/3715014.3722066 · Cited by 6
“Although our results shed light on the domain shift effects for established methods such as SimCLR, CPC, and multi-task learning, the behavior of numerous other self-supervised learning methods [8, 9, 15, 22, 42, 66] under domain shifts remains unexplored.”
Self-citationTerraSLAM: Towards GPS-Denied Localization
Xu, J., Bala, M., Eiszler, T., Chen, X., Dong, Q., Chanana, A., Pillai, P. S., & Satyanarayanan, M. (2025) · International Conference on Mobile Systems, Applications, and Services · doi:10.1145/3711875.3729144 · Cited by 2
Experience Paper: Nationwide Human Behavior Sensing in Last-mile Delivery
Hong, Z., Wang, W., Yu, A., Zhong, S., Wang, H., Ding, Y. C., He, T., & Zhang, D. (2025) · ACM Annual International Conference on Mobile Computing and Networking (MobiCom) · doi:10.1145/3680207.3765239 · Cited by 5
Liu, Z., & Fang, Q. (2025) · Discover Applied Sciences · doi:10.1007/s42452-025-07958-3 · Cited by 1
Hou, Z., Han, J., Wu, S., Wang, Z., Zhang, M., & Xu, D. (2026) · Annals of the New York Academy of Sciences · doi:10.1111/nyas.70222 · Cited by 1
“Prior to pretraining, the raw data are normalized to keep the values of different channels within a consistent range, ensuring stable and effective model training [ 17 ].”
Chen, Q., Wang, X., Chen, R., Hua, S., Li, Y., Liu, S., & Zhan, Y. (2026) · Sensors · doi:10.3390/s26082314 · Cited by 1
Multi-modal semi-supervised approach to domain adaptive human activity recognition
Noormohammadzadehmaleki, D., Oskouei, M. B., Farhadi, A., & Zamanifar, A. (2026) · Discover Computing · doi:10.1007/s10791-026-10279-7 · Cited by 1
“Inspired by foundation models in NLP and vision [40], several groups now pre-train general-purpose inertial encoders on huge, heterogeneous datasets (different users, devices, activities).”
InfantMotion2Vec: Unlabeled Data-Driven Infant Pose Estimation Using a Single Chest IMU
Khan, M. N. H., McElwain, N. L., Hasegawa‐Johnson, M., & Islam, B. (2024) · Wearable and Implantable Body Sensor Networks · doi:10.1109/bsn63547.2024.10780750 · Cited by 1
Saga: Capturing Multi-granularity Semantics from Massive Unlabelled IMU Data
Li, Y., Hu, F., Zhu, H., Zhang, S., Zhang, L., Chang, S., & Guo, M. (2025) · IEEE International Conference on Distributed Computing Systems (ICDCS) · doi:10.1109/icdcs63083.2025.00090 · Cited by 1
“As a result, we choose to use Span Masking [18], [44] to mask continuous IMU data points.”
InfluentialOn Network-Efficient Multimodal Multi-Vantage Foundation Models for Distributed Sensing
Wang, T., Chen, Y., Zhao, H., Lyu, Y., Li, J., Kimura, T., Hu, Y. C., Kara, D., Wigness, M., Twigg, J. N., et al. (2025) · IEEE International Conference on Mobile Ad-Hoc and Sensor Systems (MASS) · doi:10.1109/mass66014.2025.00017 · Cited by 2
“LLM-inspired approaches such as LIMU-BERT [20], IoT-LM [21], LLMSense [22], and Penetrative AI [23] enable zero-/few-shot inference via prompting, summarization, or adapters.”
UniMTS: Unified Pre-training for Motion Time Series
Zhang, X., Teng, D., Chowdhury, R. R., Li, S., Hong, D., Gupta, R. K., & Shang, J. (2024) · arXiv · Cited by 9
“We simulate motion time series from motion skeleton data (Section 3.1) and augment them for orientation generalization (Section 3.2).”
InfluentialFusion of Learned Representations for Multimodal Sensor Data Classification
Hinkle, L. B., Atkinson, G., & Metsis, V. (2023) · IFIP advances in information and communication technology · doi:10.1007/978-3-031-34111-3_34
“LIMU-BERT [22] builds upon the concepts of the natural language model BERT to process inertial measurement unit (motion) data for HAR.”
ViFiT: Reconstructing Vision Trajectories from IMU and Wi-Fi Fine Time Measurements
Cao, B. B., Alali, A., Liu, H., Meegan, N., Gruteser, M., Dana, K., Ashok, A., & Jain, S. (2023) · ACM MobiCom Workshop on Integrated Sensing and Communications Systems (ISACom) · doi:10.1145/3615984.3616503 · Cited by 2
“. . .”
BackgroundMethodInfluentialMSIN: An Efficient Multi-head Self-attention Framework for Inertial Navigation
Shi, G., Pan, B., & Ni, Y. (2024) · Lecture notes in computer science · doi:10.1007/978-981-97-0834-5_26
On the Effectiveness of Training Objectives of Pretrained Models for Inertial Sensor Data*
Quinlan, P., Li, Q., & Zhu, X. (2024) · Conference proceedings - Canadian Conference on Electrical and Computer Engineering · doi:10.1109/ccece59415.2024.10667341
“If masking is successful, all masked values are replaced by zeros, which was first introduced in [16].”
BackgroundMethodInfluentialSGSM: A Foundation-model-like Semi-generalist Sensing Model
Yang, T., Zhou, H., Liu, S., Guo, K., Hou, Y., Du, H., Liu, Z., & Li, X. (2024) · International Workshop on Quality of Service · doi:10.1109/iwqos61813.2024.10682922
“However, the current cost of designing intelligent sensing systems is relatively high since the models were designed to solve specific tasks with expensive expert knowledge [7] or a substantial amount of domain-specific data [8], one at a time.”
BackgroundSGSM: semi-generalist sensing model combining handcrafted and deep learning methods
Yang, T., Zhou, H., Liu, S., Guo, K., Hou, Y., Du, H., & Li, X. (2024) · International Journal of Machine Learning and Cybernetics · doi:10.1007/s13042-024-02396-w
“LIMU-BERT [8] adopts the natural language model BERT to capture temporal relations and feature distributions in IMU sensor measurements.”
InfluentialFast Adapting Few-Shot Federated Learning System for Human Activity Recognition
Zhang, Z., & Zhao, Y. (2024) · IEEE International Symposium on Parallel and Distributed Processing with Applications (ISPA) · doi:10.1109/ispa63168.2024.00264 · Cited by 2
Synthetic Sensory Data Generators: How Much Progress Are We Making?
Sharma, R., Khamis, A., Khalifa, S., Bretherton, D., & Kusý, B. (2025) · IEEE International Conference on Pervasive Computing and Communications Workshops (PerCom Workshops) · doi:10.1109/percomworkshops65533.2025.00084 · Cited by 1
“This raises fundamental questions about whether SSDGs provide any advantage over alternative approaches such as few-shot learning [20], [21], which achieves competitive results without the need for synthetic data generation.”
Guo, Y., Huang, Q., & Chen, X. (2025) · International Conference on Electronics, Circuits and Information Engineering (ECIE) · doi:10.1109/ecie65947.2025.11087045
“The advancement of sensing devices has driven the development of intelligent sensing technologies, including human activity recognition [3] and health monitoring [4].”
Inertial Signal Forecasting With Foundation Model Techniques
Wieland, C., & Pankratius, V. (2025) · IEEE Sensors Journal · doi:10.1109/jsen.2025.3603456 · Cited by 3
“Masking approaches, such as span masking [30] or spatial masking [35], learn to reconstruct signals with masked values.”
Indoor Localization from Large-Scale Poor-Quality Crowdsourcing Wifi Data for On-Demand Delivery
Guo, K., Zheng, S., Zhou, H., Zhang, Y., Yan, K., Shen, G., Du, H., & Li, X. (2025) · International Workshop on Quality of Service · doi:10.1109/iwqos65803.2025.11143426
“LIMU-BERT [19] modifies the BERT architecture to adapt to IMU data and achieves promising results in human activity recognition tasks.”
Self-citationBuilding Generalizable Deep Learning Solutions for AIoT Applications
Xu, H. (2025) · International Conference on Mobile Systems, Applications, and Services · doi:10.1145/3711875.3736673
Self-citationData-Free Backdoor Attacks on Self-Supervised Human Activity Recognition Models
Zhao, T., & Wang, X. (2025) · IEEE International Conference on Mobile Ad-Hoc and Sensor Systems (MASS) · doi:10.1109/mass66014.2025.00051
“Consequently, several SSL methods have been explored for various IoT sensing tasks [2].”
Experience Paper: Adopting Activity Recognition in On-demand Food Delivery Business
Xu, H., Zhang, Y., Gao, W., Shen, G., & Li, M. (2025) · ACM Annual International Conference on Mobile Computing and Networking (MobiCom) · doi:10.1145/3680207.3765261 · Cited by 5
Self-citationZhang, Z., Zhao, D. Y., Zhu, G., Li, X. Q., Li, C., Liu, B., Li, Y., & Ma, H. (2026) · IEEE Transactions on Mobile Computing · doi:10.1109/tmc.2026.3672513 · Cited by 1
“Although some works [12], [18], [19], [20] claim to build a foundation HAR model, they either cannot be applied to multi-modal data [12], [18], [20] or have limitations in combining multiple heterogeneous datasets during pre-training [19].”
InfluentialArmPad: Transforming Forearms Into Interaction Interfaces With Smartwatches
Yang, Z., Yang, Q., Liu, Z., Li, Z., Zou, Y., & Wu, K. (2026) · IEEE Transactions on Mobile Computing · doi:10.1109/tmc.2026.3672486
“Pre-trained models capture trans-ferable motion patterns (e.g., acceleration decay, spectral features), enhancing robustness to noise [49].”
EchoScriptor: Automatic Lifelogging Narratives via Activity-Based Audio–Language Model
Li, K. Y., Zhou, X., Zheng, H., Shin, K. G., & Sample, A. P. (2026) · CHI Conference on Human Factors in Computing Systems · doi:10.1145/3772318.3791528 · Cited by 1
“…studied in mobile systems, wearable technology, healthcare, and ubiquitous computing, with modalities such as cameras [57, 81], IMUs [7, 43, 95, 96, 98], microphones [54, 62], GPS [30, 94], and RF signals [13, 51], each presenting distinct trade-offs in coverage, reliability, and practicality.”
Physical Self-Supervised Learning: IMU Sensing without Manual Labels
Leng, Y., Liu, R., Hu, S., Zhao, P., Chen, C.-F., Chen, S., & Yao, S. (2026) · International Conference on Mobile Systems, Applications, and Services · doi:10.1145/3745756.3809252 · Cited by 2
“• LIMU-BERT [70]:Self-supervised BERT-like architecture with pre-training and fine-tuning.”
InfluentialIMU2Video: World Model-Enabled Video Generation from IMU sensors
Li, Q., Wu, R., & Xu, W. (2026) · International Conference on Mobile Systems, Applications, and Services · doi:10.1145/3812836.3814749
From IMU Streams to Real-Time Decisions: Past-Only Next-Window Badminton Action Prediction
Zhu, Q., Wang, J., & Guo, B. (2026) · Sensors · doi:10.3390/s26123651 · Cited by 1
“To generate reliable labels for the IMU sensor data of badminton players, we adopt a self-supervised learning-based annotation method, which is derived from the LIMU-BERT framework proposed by Xu et al. [21].”
Evaluating Visual Prompts with Eye-Tracking Data for MLLM-Based Human Activity Recognition
Choi, J. Y., Kim, S. G., Yoon, H., Lee, T., Lee, D., Chung, J., Kil, J., Rossi, R., Lee, S.-J., & Lee, T. Y. (2026) · IEEE Pacific Visualization Conference (PacificVis) · doi:10.1109/pacificvis68791.2026.00023
“…have focused on aligning time-series modalities with LLMs, either by developing specialized tokenizers and encoders to handle numerical sequences efficiently [9] or by training foundation models directly on large-scale time-series datasets such as inertial measurement unit (IMU) data [10, 11].”
Wang, G. Z., Yang, S., Song, R., Zhang, D., Sun, Q., & Chen, Y. (2026) · IEEE Transactions on Mobile Computing · doi:10.1109/tmc.2026.3711490
Diffusion denoiser-aided gyrocompassing
Ben-Arie, G., Engelsman, D., Dror, R., et al. (2026) · Measurement · doi:10.1016/j.measurement.2026.122370 · Cited by 1
“In parallel, self-supervised learning approaches have emerged as an attractive framework for inertial sensor processing by exploiting large quantities of unlabeled IMU data to learn robust motion representations and improve sensor signal quality [38].”
Wang, B., Zhang, Y., Yu, J., & Yang, J. (2026) · IEEE Internet of Things Journal · doi:10.1109/jiot.2026.3711146 · Cited by 1
“In inertial sensing, frameworks such as SelfHAR [29] and LIMU-BERT [30] pretrain on large-scale unlabeled IMU sequences to capture temporal dependencies, but these motion-specific approaches are less applicable to device-free modalities like acoustic or RF signals.”
GenHAR: Generalizing Cross-domain Human Activity Recognition for Last-mile Delivery
Hong, Z., Li, Z., Fan, X., Yang, G., Guo, B., Wang, H., He, T., & Zhang, D. (2026) · ACM SIGKDD Conference on Knowledge Discovery and Data Mining (KDD) · doi:10.1145/3770854.3783921 · Cited by 2
UniMoRe : Uni fied Mo tion Re presentation via Gait IMU Pretraining
Chakraborty, S., Kumar, A., & Nigam, A. (2026) · IEEE Sensors Journal · doi:10.1109/jsen.2026.3718949
“The early works introduced encoder-based masked reconstruction for IMU self-supervision, preceding widespread transformer adoption [24], [25].”
InfluentialCalatrava-Nicolás, F. M., Stoyanov, T., & Martinez Mozos, O. (2026) · Scientific Reports · doi:10.1038/s41598-026-64846-5
“SSL methods for HAR exploit a variety of proxy objectives to learn from unlabeled data, including transformation recognition 25 , temporal predictive coding 22,38 , contrastive view agreement 26,39 , non-contrastive view agreement 27,28 , and masked reconstruction 23,24,40 among others.”
Shift-Aware Operator Selection for Label-Free Test-Time Calibration in Cross-Domain Dataset
Deshagouni, M., & Akther, S. (2026) · IEEE International Conference on Big Data Computing Service and Applications (BigDataService) · doi:10.1109/bigdataservice70481.2026.00051
“Deep HAR has evolved from CNNs [8] and DeepConvLSTM [6] to attention [9] and self-supervised models [20].”
Revisiting Time-Domain Interpretability for Self-Supervised IMU Sensing Models
Wang, Y., Zhao, T., & Wang, X. (2026) · IEEE/ACM Conference on Connected Health: Applications, Systems and Engineering Technologies · doi:10.1109/chase69719.2026.00039
“We further consider two representative generative SSL methods: LIMU-BERT [20], which learns contextual sensor representations through masked reconstruction, and FLM [21], which adapts frozen pre-trained language models for time-series learning and serves as a strong large-scale pretraining baseline.”
Yang, Q., Huang, Y., Han, J., Liang, H., Wei, H., & Hou, Z. (2026) · Processes · doi:10.3390/pr14182876
Adapting Pretrained Large Vision Models for Sensor-based Activity Recognition
Cai, Y., Feng, R., Cai, K., Liu, Y., Guo, B., & Hong, Z. (2026) · Proceedings of the ACM on Interactive Mobile Wearable and Ubiquitous Technologies · doi:10.1145/3832003
“We preprocess all sensor data to a unified sampling rate of 20 Hz and segment them into fixed-length windows of 6 seconds [120].”
InfluentialFM-FoG: A Real-Time Foundation Model-based Wearable System for Freezing-of-Gait Mitigation
Chi, C., Clapham, J., Cloud, L. J., Pretzer‐Aboff, I., Blackwell, G. G., Shao, H., & Zhou, G. (2026) · Proceedings of the ACM on Interactive Mobile Wearable and Ubiquitous Technologies · doi:10.1145/3831651 · Cited by 1
Haresamudram, H., Tang, C. I., Suh, S., Lukowicz, P., & Ploetz, T. (2024) · arXiv · Cited by 31
“…recognition is yet to be verified, these works provide a good vision of how foundational models for human activity recognition can be: that training with a well-selected set of datasets and modest model size could offer much more practical and effective solutions for this mobile sensing task.”
Metadata
- Authors
- Xu, H., Zhou, P., Tan, R., Li, M., & Shen, G.
- Affiliations
- Nanyang Technological University (SG)
- Alibaba Group (China) (CN)
- Venue
- Proceedings of the 19th ACM Conference on Embedded Networked Sensor Systems
- Publisher
- ACM
- Published
- 2021-11-15
- Pages
- 220-233
- OpenAlex
- W3211771663
- Cited by
- 150
- References
- 56
- Citations by year
- 2026: 28 · 2025: 48 · 2024: 47 · 2023: 22 · 2022: 5
- FWCI
- 6.04
- Type
- conference paper
- Language
- en
- Open access
- closed
- Locations
- ACM Conference on Embedded Networked Sensor Systems (SenSys) (publishedVersion)
- Rare & Special e-Zone (The Hong Kong University of Science and Technology) (submittedVersion)
- Funders
- NTU CoE SUG
- National Research Foundation, Singapore under its Industry Alignment Fund ? Pre-positioning (IAF-PP) Funding Initiative
- Alibaba Group through Alibaba Innovative Research (AIR) Program and Alibaba-NTU Singapore Joint Research Institute (JRI)
- Topics
- Physical Sciences › Computer Science › Computer Vision and Pattern Recognition › Context-Aware Activity Recognition Systems
- Physical Sciences › Engineering › Aerospace Engineering › Inertial Sensor and Navigation
- Physical Sciences › Computer Science › Artificial Intelligence › Domain Adaptation and Few-Shot Learning
- Keywords
- inertial measurement unit, self-supervised learning, representation learning, mobile devices, human activity recognition, lightweight models, limited labeled data
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
Data sources
Bibliographic data: Semantic Scholar, OpenAlex, Crossref.
Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).
Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.