Curriculum learning
The DOI matches its Crossref record; 35 of the 40 references it cites were found in a registry or an index.
Abstract
Humans and animals learn much better when the examples are not randomly presented but organized in a meaningful order which illustrates gradually more concepts, and gradually more complex ones. Here, we formalize such training strategies in the context of machine learning, and call them "curriculum learning". In the context of recent research studying the difficulty of training in the presence of non-convex training criteria (for deep deterministic and stochastic neural networks), we explore curriculum learning in various set-ups. The experiments show that significant improvements in generalization can be achieved. We hypothesize that curriculum learning has both an effect on the speed of convergence of the training process to a minimum and, in the case of non-convex criteria, on the quality of the local minima obtained: curriculum learning can be seen as a particular form of continuation method (a general strategy for global optimization of non-convex functions).
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
“Moreover, the lack of acoustical signals could even account for spatial disorientation 91,92.”
From The impact of multisensory integration and perceptual load in virtual reality settings on performance, workload and presence (2021) · cites this paper as [92]
In the citing paper
“This suggests that curriculum-style learning (Bengio et al., 2009), where simpler or archetypal problems are presented first to build a repository of valid heuristics, may potentially bootstrap performance.”
From Dynamic Cheatsheet: Test-Time Learning with Adaptive Memory (2025) · cites this paper as Bengio et al., 2009
Citations
Cites
Reducing the dimensionality of data with neural networks
Hinton, G. E., & Salakhutdinov, R. R. (2006) · Science · doi:10.1126/science.1127647 · Cited by 20,928
A fast learning algorithm for deep belief nets
Hinton, G. E., Osindero, S., & Teh, Y.-W. (2006) · Neural Computation · doi:10.1162/neco.2006.18.7.1527 · Cited by 16,336
Learning Deep Architectures for AI
Bengio, Y. (2009) · Foundations and Trends® in Machine Learning · doi:10.1561/2200000006 · Cited by 9,239
Self-citationExtracting 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
Self-citationA unified architecture for natural language processing: deep neural networks with multitask learning
Collobert, R., & Weston, J. (2008) · Proceedings of the 25th international conference on Machine learning - ICML '08 · doi:10.1145/1390156.1390177 · Cited by 5,058 · cited 6×
“In Collobert and Weston (2008), the main objective is to learn an embedding for the words as a side-e.ect of learning to compute this score.”
BackgroundMethodInfluentialSelf-citationLearning Deep Architectures for AI
Bengio, Y. (2009) · now publishers, Inc. eBooks · doi:10.1561/9781601982957 · Cited by 4,941
Self-citationGreedy layer-wise training of deep networks
Bengio, Y., Lamblin, P., Popovici, D., & Larochelle, H. (2007) · Advances in Neural Information Processing Systems 19 · doi:10.7551/mitpress/7503.003.0024 · Cited by 4,841
Self-citationA Neural Probabilistic Language Model
Bengio, Y., Ducharme, R., Vincent, P., Jauvin, C., Jaz, K., Hofmann, T., Poggio, T., & Shawe-Taylor, J. S. (2000) · Cited by 3,479
Self-citationLearning and development in neural networks: the importance of starting small
Elman, J. L. (1993) · Cognition · doi:10.1016/0010-0277(93)90058-4 · Cited by 1,988 · cited 8×
“The idea of training a learning machine with a curriculum can be traced back at least to Elman (1993).”
BackgroundMethodInfluentialRestricted Boltzmann machines for collaborative filtering
Salakhutdinov, R., Mnih, A., & Hinton, G. (2007) · Proceedings of the 24th international conference on Machine learning · doi:10.1145/1273496.1273596 · Cited by 1,885
Research works printed but not linked: 3
- 10.5555/89603 · doi:10.5555/89603
- 10.5555/866740 · doi:10.5555/866740
- 10.5555/525571 · doi:10.5555/525571
Cited by
Items 3,701–3,800 of 5,240
Rising From Pieces: Effective Inference at the Edge via Robust Split ML
Weng, Y., Zheng, T., Chen, Z., Hu, M., & Luo, J. (2025) · IEEE Transactions on Mobile Computing · doi:10.1109/tmc.2025.3605382 · Cited by 3
“Curriculum learning [49], [50] is an effective paradigm in machine learning that can address this challenge: it involves starting with simpler tasks and incrementally increasing the complexity, ensuring a stable training process.”
Adaptive formation control and transformation in virtual crowds via deep reinforcement learning
Sun, L., Qiu, Y., & Qin, W. (2025) · The Visual Computer · doi:10.1007/s00371-025-04163-6 · Cited by 1
MCTS-Based Policy Improvement for Reinforcement Learning
Csippán, G., Péter, I., Kővári, B., & Bécsi, T. (2025) · Machine Learning and Knowledge Extraction · doi:10.3390/make7030098 · Cited by 2
“The authors in [15] introduced Curriculum Learning, which organizes training samples to gradually increase task difficulty, helping the model to improve its performance on complex tasks.”
Yi, L., Qin, Y., & Huang, R. (2025) · IEEE Transactions on Multimedia · doi:10.1109/tmm.2025.3604930 · Cited by 3
“Restrictions apply. strategy [20], enabling RL models to perform better across a wider range of network environments.”
An efficient dual space curriculum learning approach for speech emotion recognition
Karani, R. B., & Desai, S. S. (2025) · Signal Image and Video Processing · doi:10.1007/s11760-025-04796-z · Cited by 1
“Curriculum Learning (CL), introduced by Bengio et al. [7], improves convergence and optimization across tasks such as toy problems, NLP, and categorization.”
Lanzoni, D., Montalenti, F., & Bergamaschini, R. (2025) · Journal of Physics Condensed Matter · doi:10.1088/1361-648x/ae096d · Cited by 7
“The first is to progressively expand the time window learned by the PINN (an approach resembling curriculum learning [96]).”
Mix-Based Training Strategies for Learning Implicit Neural Representations
Han, D., Zhang, C., Zheng, S., et al. (2025) · IEEE Transactions on Multimedia · doi:10.1109/tmm.2025.3613162 · Cited by 1
“Curriculum Learning, as proposedbyBengioet al. [37], draws inspiration from both human and animal learning processes.”
Park, Y. H., Choi, H., Lee, K.-B., & Kim, H. (2025) · Applied Sciences · doi:10.3390/app151910337 · Cited by 2
Duminil, A., Ieng, S.-S., & Gruyer, D. (2025) · Scientific Reports · doi:10.1038/s41598-025-15480-0 · Cited by 1
“To achieve the building of the models, the curriculum learning approach 33 is used.”
Ciubotaru, B.-I. (2025) · International Conference on Control Systems and Computer Science · doi:10.1109/cscs66924.2025.00063 · Cited by 3
“Innovations in training methodologies aim to reduce computational and data requirements through advances in self-supervised learning, curriculum learning, continual learning, and federated learning [60].”
Few-Shot Knowledge Graph Completion With Adaptive Negative Sampling Mechanism
Yu, J., Rui, L., Lin, Y., Gao, Z., Qiu, X., & Guo, S. (2025) · IEEE Transactions on Knowledge and Data Engineering · doi:10.1109/tkde.2025.3614171 · Cited by 1
“Curriculum learning (CL) [35] is a training strategy which mimics the sequence of learning that is effective in human education.”
Overcoming Data Scarcity: Brain Tumor Segmentation in Pediatric and African Populations
Al-Fakih, A., Rezk, A. M., Shazly, A., Ryu, K., & Al‐masni, M. A. (2025) · Lecture notes in computer science · doi:10.1007/978-3-032-08009-7_2 · Cited by 1
Singh, R., & Guggilam, S. (2025) · IEEE Access · doi:10.1109/access.2025.3621553 · Cited by 1
“While various methods enhance neural network training, curriculum learning [3], [31], [27] has garnered significant interest in recent years.”
Dynamic deep multi-label image data augmentation based on self-paced learning
Yu, B., Li, W., Zhang, C., et al. (2025) · Computer Vision and Image Understanding · doi:10.1016/j.cviu.2025.104530 · Cited by 1
A Readability-Driven Curriculum Learning Method for Data-Efficient Small Language Model Pretraining
Kim, S., Park, J., & Kim, J. (2025) · Mathematics · doi:10.3390/math13203300 · Cited by 1
Skill-Nav: enhanced navigation with versatile quadrupedal locomotion via waypoint interface
Wang, D., Bai, C., Li, C., Shi, J., Ding, Y., Zhang, C., & Zhao, B. (2025) · Vicinagearth. · doi:10.1007/s44336-025-00015-y · Cited by 7
“Reinforcement Learning (RL)-based methods [6–9], trained in simulated environments like Isaac Gym [10], utilize curriculum learning [11] and domain randomization to enhance robustness and versatility.”
Mu, B., Chen, C., Yin, K., Yang, L., & Xia, Y. (2025) · Control Engineering Practice · doi:10.1016/j.conengprac.2025.106622 · Cited by 1
Emotion Recognition in Multi-speaker Scenarios for Social Robots
García‐Rodríguez, J., Begazo, R., Cárdenas, P., Dongo, I., Cardinale, Y., & Aguilera, A. (2025) · International Journal of Social Robotics · doi:10.1007/s12369-025-01312-8 · Cited by 2
“This reasoning is based on the concept of curriculum learning [69 ], where training on a more difficult task helps models generalize better to easier conditions.”
PMG: Progressive Motion Generation via Sparse Anchor Postures Curriculum Learning
Xi, Y., Zhang, J., & Yang, X. (2025) · ACM International Conference on Multimedia (ACM MM) · doi:10.1145/3746027.3754866 · Cited by 2
“To alleviate this challenge, we introduce a novel training strategy, termed Sparse Anchor Posture Curriculum Learning (SAP-CL) [3, 29, 31, 35, 36, 43, 49, 57, 60, 62].”
Suhas, M., Abisset‐Chavanne, E., & Rey, P.-A. (2025) · Applied Sciences · doi:10.3390/app152111532 · Cited by 2
Towards Aligned Data Forgetting via Twin Machine Unlearning
Ji, H., Zheng, L., Sun, Y., Gao, F., Wang, Y., Gao, H., & Niu, Z. (2025) · IEEE International Conference on Multimedia and Expo (ICME) · doi:10.1109/icme59968.2025.11208918 · Cited by 2
“Curriculum learning theory [14], [15] suggests that easy samples are learned earlier than hard samples.”
High-Context Empathy in Conversations for Large Language Models
Chen, Y., Xia, L., Cao, J., Hou, Z., Dai, W., & Li, Z. (2025) · ACM International Conference on Information and Knowledge Management (CIKM) · doi:10.1145/3746252.3761424 · Cited by 2
“The third step is Empathetic-oriented Meta-Curriculum Learning (EMCL) which is to automatically learn a best sequence of different tasks generally from easiest to hardest based on Meta Learning (ML) [19] and Curriculum Learning (CL) [4].”
cMALC-D: Contextual Multi-Agent LLM-Guided Curriculum Learning with Diversity-Based Context Blending
Satheesh, A., Powell, K., & Wei, H. (2025) · ACM International Conference on Information and Knowledge Management (CIKM) · doi:10.1145/3746252.3760863 · Cited by 6
“Curriculum learning [1] aims to improve generalization by structuring agents’ learning process, progressing from easier to more challenging tasks.”
MedFormer: a data-driven model for forecasting the Mediterranean Sea
Epicoco, I., Donno, D., Accarino, G., Norberti, S., Grandi, A., Giurato, M., McAdam, R., Elia, D., Clementi, E., Nassisi, P., et al. (2026) · Earth Science Informatics · doi:10.1007/s12145-026-02169-9 · Cited by 3
“During this phase, training is performed using higher quality datasets: the very high resolution Mediterranean analysis, from the MedFS system [9], and surface meteorological forcing from the ECMWF HRES (high resolution) operational analysis at 1 / 10 ◦ .”
Task-adaptive debiasing with SCM for sentiment analysis
Zhu, C., Ziabari, S. S. M., & Alsahag, A. M. M. (2025) · Machine learning for computational science and engineering · doi:10.1007/s44379-025-00043-x · Cited by 4
“√ Our rank schedule follows the easy-to-hard paradigm of Curriculum Learning [33] and Self-Paced Learning [34].”
An Easy-to-Hard Curriculum Learning Framework for Fast Production Cost Minimization Simulation
Guo, Z., Qu, C., Qian, T., Hu, Q., Hu, R., Wang, Y., Liu, H., Yu, W., & Alhazmi, M. (2025) · IEEE Transactions on Smart Grid · doi:10.1109/tsg.2025.3632361 · Cited by 1
“The advantages of CL, such as improved convergence speed and higher quality of local minima, are highlighted in [28].”
Mridul, M. A., & Seneviratne, O. W. (2025) · ACM International Conference on AI in Finance (ICAIF) · doi:10.1145/3768292.3770397 · Cited by 2
“To address the inherent sparse reward problem in code synthesis, where most randomly generated contracts are invalid, we employ a two-phase curriculum learning strategy [3].”
Research on Dynamic Curriculum Learning in Mongolian-Chinese Neural Machine Translation
Hu, C., Si, Q., & Wang, S. (2025) · doi:10.1109/ijcnn64981.2025.11228447 · Cited by 1
“Additionally, we introduce Curriculum Learning(CL) [3] to optimize data utilization in low-resource settings.”
Choi, J.-W., Rhee, W. L., Han, D.-H., et al. (2025) · Diagnostics · doi:10.3390/diagnostics15232951 · Cited by 1
Pathare, D., Laine, L., & Chehreghani, M. H. (2025) · Artificial Intelligence Review · doi:10.1007/s10462-025-11448-8 · Cited by 4
“…training of RL agent with the complex TCOP based reward function consisting of realistic costs and revenue values with and without normalization (ii) extend our RL framework by incorporating curriculum learning techniques [25, 26] and compare the results with the non-curriculum learning approach.”
CL-DRW: Curriculum Learning-Based Deep Robust Watermarking for Social Networks
Zhao, Y., Zhu, G., Zhou, J., Li, X., Zhang, H., & Zhang, X. (2025) · IEEE Transactions on Circuits and Systems for Video Technology · doi:10.1109/tcsvt.2025.3637537 · Cited by 1
“Forthe training scheduler, we adopt a curriculum learning approach known as “Baby Step” [37].”
Fang, X., Tao, L., & Li, Z. (2025) · IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) · doi:10.1109/iros60139.2025.11247170 · Cited by 1
“Curriculum learning has shown significant promise in enabling agents to tackle progressively more challenging tasks by structuring the training process in incremental stages [4], [2].”
RoCaRS: Robust Camera-Radar BEV Segmentation for Sensor Failure Scenarios
Park, B., Kim, J., Cho, Y., & Hwang, S. (2025) · IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) · doi:10.1109/iros60139.2025.11246952 · Cited by 2
“We leverage a method that increases the dropout rate gradually by iterations, inspired by curriculum learning [24].”
PC 2 P: Multi-Agent Path Finding via Personalized-Enhanced Communication and Crowd Perception
Li, G., Xu, S., Hao, Y., et al. (2025) · IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) · doi:10.1109/iros60139.2025.11247630 · Cited by 1
VertiSelector: Automatic Curriculum Learning for Wheeled Mobility on Vertically Challenging Terrain
Tong mo, X. U., Pan, C., & Xiao, X. (2025) · IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) · doi:10.1109/iros60139.2025.11246235 · Cited by 7
“This idea was expanded upon by Bengio et al. [49], who proposed a learning paradigm where training examples are presented in a meaningful order, gradually increasing in complexity.”
GACL: Grounded Adaptive Curriculum Learning with Active Task and Performance Monitoring
Wang, L., Xu, Z., Stone, P., & Xiao, X. (2025) · IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) · doi:10.1109/iros60139.2025.11246910 · Cited by 1
“Curriculum learning, originally inspired by the way humans acquire skills, has been explored in robotics to structure learning from increasingly complex tasks [8].”
Analyzing homogenous and heterogeneous multi-server queues via neural networks
Sherzer, E. (2025) · Annals of Operations Research · doi:10.1007/s10479-025-06974-1 · Cited by 1
Wang, W., Zhu, L., Liu, Z., Zhang, J., Chen, Y., & Gu, Y. (2025) · IEEE Transactions on Cognitive Communications and Networking · doi:10.1109/tccn.2025.3642815 · Cited by 1
“To overcome this, recent research has proposed curriculum learning (CL) strategies [33], [34], where models are trained with adversarial inputs of gradually increasing intensity.”
Jung, S., Jeoung, J., Kong, M., Jung, D., & Hong, T. (2025) · Advanced Engineering Informatics · doi:10.1016/j.aei.2025.104220 · Cited by 3
Liu, Z., Zeng, K., Ma, Q., & Kwok, J. T.-Y. (2025) · IEEE Transactions on Pattern Analysis and Machine Intelligence · doi:10.1109/tpami.2025.3644603 · Cited by 7
“Furthermore, DNNs may exhibit varying degrees of learning difficulties across different categories within a dataset [50].”
DPARNet-RSE: Toward Angular Region-Customizable Speech Extraction
Yang, Y., & Zhou, C. (2025) · IEEE Signal Processing Letters · doi:10.1109/lsp.2025.3613271 · Cited by 1
“To address this issue, we propose a two-stage training strategy inspired by curriculum learning concept [24] and its successful applications in speech front-end tasks [25], [26].”
Liang, Y., Wu, J., Zhang, J., Chang, Y., Feng, J., & Shi, G. (2025) · IEEE Transactions on Geoscience and Remote Sensing · doi:10.1109/tgrs.2025.3643494 · Cited by 1
“Drawing on the principle of learning from easy to hard examples, curriculum learning [39], [40] sequentially organizes training samples by increasing di ffi culty to facilitate model convergence and improve generalization.”
Routing-Aware RL for Mobile Relay Navigation in Disconnected Ad Hoc Networks
Karacelebi, C., Onur, E., & Şahin, Y. (2025) · International Conference on Network and Service Management (CNSM) · doi:10.23919/cnsm67658.2025.11297454 · Cited by 1
“Such level-resampling curricula are common in RL to promote generalization [8]–[10].”
Plant growth point localization via epoch-based prior annealing
Ma, C., Zhang, Z., Tian, F., et al. (2026) · Scientific Reports · doi:10.1038/s41598-026-35009-3 · Cited by 1
“Curriculum learning 29 is a training strategy inspired by human learning, in which a model is first exposed to simple tasks and data before gradually progressing to more complex ones.”
Prasasta, R. S., Pradata, S. Y., Santiyuda, G., Amrizal, M. A., Pulungan, R., & Takizawa, H. (2026) · doi:10.1145/3784828.3785163 · Cited by 1
“Building upon this, Budiarjo et al. [9] integrated curriculum learning (CL) [7] to accelerate policy convergence in nodes’ on/off decision-making.”
A Survey on Video Captioning in the Era of Large Language Models
Brimont, A., Zaharia, T. B., & Țapu, R. (2026) · ACM Transactions on Multimedia Computing Communications and Applications · doi:10.1145/3793908 · Cited by 5
“In this context, selective training [49], where only specific modules are fine-tuned, and curriculum learning [10], where the training task difficulty is gradually increased, have become widely adopted practices.”
Path following and stabilization of a bicycle model using a reinforcement learning approach
Weyrer, S., Manzl, P., Schwab, A. L., & Gerstmayr, J. (2026) · Multibody System Dynamics · doi:10.1007/s11044-026-10144-x · Cited by 3
“Curriculum Learning (CL) is a training strategy for the field of machine learning that was first formalized in the work [36].”
Fine-Grained Alignment Supervision Matters in Vision-and-Language Navigation
He, K., Huang, Y., Jing, Y., Wu, Q., & Wang, L. (2026) · IEEE Transactions on Pattern Analysis and Machine Intelligence · doi:10.1109/tpami.2026.3658949 · Cited by 8
“Toclearlyhighlighttheeffectiveness ofthefine-graineddataprovidedbyLandmark-RxR,weadopt thenaivecurriculumlearningstrategy[22]asourapproach.”
Simulation Based Reward Function Validation for Multi-Agent On Orbit Inspection
Quinn, P. D., Gopu, B. P. R., Nehma, G., & Tiwari, M. (2026) · AIAA SciTech Forum · doi:10.2514/6.2026-2042 · Cited by 1
Li, W., Liang, S., Chen, Y., Ma, H., Xu, J., Ma, Y., Chen, Z., Fang, H., & Zhang, F. (2026) · IEEE Journal of Selected Topics in Signal Processing · doi:10.1109/jstsp.2026.3660045 · Cited by 3
“The correction process is guided by the paradigm of curriculum learning [63], [64], which follows the criterion that learns from easytohard.”
A parametrically‐Conditioned Deep Learning Surrogate for Coherent Spinodal Decomposition
Fantasia, A., Lanzoni, D., Eugenio, N. D., Monteleone, A., Bergamaschini, R., & Montalenti, F. (2026) · Advanced Theory and Simulations · doi:10.1002/adts.202502144 · Cited by 2
“Thanks to the recurrent character of the model, the so-called curriculum learning technique [ 63 ] has been exploited.”
Curriculum Learning With Heat Map-Based State Representations in Autonomous Drone Cargo Search
Kömürcü, K. K., Kafadar, K. D., Ozaltun, E., Orak, F., Şanlı, M. E., Gazi, E., & Üre, N. K. (2026) · AIAA SciTech Forum · doi:10.2514/6.2026-2205 · Cited by 1
Fiston, O. O., Djungu, S. J., Mbayandjambe, A. M., Mwamba, F., Biaba, J., Frey, S., Simboni, T. S., Kasoro, N. M., & Muhala, B. L. (2025) · Analecta Technica Szegedinensia · doi:10.14232/analecta.2025.3-4.15-38 · Cited by 2
Verma, N., Ranvijay, R., & Yadav, D. K. (2026) · International Journal of Imaging Systems and Technology · doi:10.1002/ima.70307 · Cited by 1
“Bengio et al. [23] proposed curriculum learning, which derived inspiration from human learning approaches.”
Peng, H., Wang, Y., Miao, H., et al. (2026) · IEEE Signal Processing Letters · doi:10.1109/lsp.2026.3663884 · Cited by 1
“More-over, the Memory-Aware Attack (MAO) determines the training sequence by ranking new representative attacks in descending feature similarity to the previous attack r o , consistent with the principles of curriculum learning [17], [18], [19].”
From Prediction to Prescription: A Causal-AI for Personalized Learning Path Optimization in MOOCs
Isiwu, C. P., Nnadi, L. C., & Watanobe, Y. (2026) · IEEE Access · doi:10.1109/access.2026.3666619 · Cited by 4
Achieving Stable High-Speed Locomotion for Humanoid Robots with Deep Reinforcement Learning
Zhang, X., Wang, X., Zhang, L., Guo, G., Shen, X., & Zhang, W. (2025) · IEEE International Conference on Robotics and Biomimetics (ROBIO) · doi:10.1109/robio66223.2025.11377915 · Cited by 12
“To address this issue, we trained the humanoid robot using low-velocity commands initially and progressively expanded the velocity range as per the curriculum [22].”
Pillai, A., Azri, N. R., Pornphol, P., Sathasivam, S., & Inthanil, A. (2026) · Franklin Open · doi:10.1016/j.fraope.2026.100543 · Cited by 2
Adapting Curricula by Tracking the Gap between Validation and Training Loss
Singh, R. (2025) · IEEE International Conference on Computer Vision Workshops · doi:10.1109/iccvw69036.2025.00541 · Cited by 1
“Curriculum learning [5] is inspired by the way humans and animals learn through structured and meaningful presentation of topics.”
Progressively Alleviating Noise for Unsupervised Cross-Domain Image Retrieval
Wang, Z., Li, Q., Xu, X., Zhu, L., Yang, Y., & Shen, H. T. (2026) · IEEE Transactions on Multimedia · doi:10.1109/tmm.2026.3664950 · Cited by 2
“Motivated by Curriculum Learning [57], we develop a progressive learning method to gradually mitigate noise and then assist the model in accurately capturing the semantics.”
CauSTream: Causal Spatio-Temporal Representation Learning for Streamflow Forecasting
Wan, S., Shah, R. D., Sabo, J. L., Liu, H., & Candan, K. S. (2025) · IEEE International Conference on Big Data (BigData) · doi:10.1109/bigdata66926.2025.11402305 · Cited by 2
“The complete model is trained end-to-end by optimizing a weighted sum of all loss components: We employ a curriculum learning strategy [25], first training on the forecast and ELBO objectives before gradually annealing the weights of the structural penalties.”
DACL-RAG: Data Augmentation Strategy With Curriculum Learning for Retrieval-Augmented Generation
Wang, S., Zhang, L., Fu, Z., Mao, Z., & Zhang, Y. (2026) · IEEE Transactions on Audio Speech and Language Processing · doi:10.1109/taslpro.2026.3675775 · Cited by 5
“Curriculum Learning (CL) [26] is a machine learning strategy that mimics human learning by training models gradually, starting with simpler tasks and progressing to more complex ones.”
RAES: a reward-aligned expert sequencing framework for long-horizon robotics
Acs, M., & Zhong, X. (2026) · International Journal of Intelligent Robotics and Applications · doi:10.1007/s41315-026-00524-z · Cited by 2
“Curriculum learning reduces exploration difficulty by struc - turing training from simpler to harder versions of the task (Bengio et al. 2009).”
Dynamic Prompt Compression for Efficient Inference of Large Language Models
Hu, J., Zhang, W., Wang, Y., Hu, Y., Xiao, B., Tan, M., & Du, Q. (2026) · IEEE Transactions on Knowledge and Data Engineering · doi:10.1109/tkde.2026.3669558 · Cited by 13
“Since its introduction [55], CL has been widely studied across different domains [56], [57], [58], [59].”
Guo, M., Huang, H., Zeng, F., Nie, M., & Dang, M. (2026) · Measurement · doi:10.1016/j.measurement.2026.121178 · Cited by 1
CAM3R: Camera-Agnostic Model for 3D Reconstruction
Guruprasad, N., Yadav, A., Chellappa, R., & Peng, C. (2026) · Lecture notes in computer science · doi:10.1007/978-3-032-37321-2_14 · Cited by 7
Fu, Y., Guo, J., Li, D., Chen, J., Han, L., Qu, Y., & Pan, Y. (2026) · Robotica · doi:10.1017/s0263574726103282 · Cited by 1
“Additionally, to accelerate training and improve the control performance of DRL, the concept of curriculum learning (CL) [4] has been integrated into RL. Narvekar [23] systematically explored the assumptions, capabilities, and goals of CL methods in RL, and Luo [20] introduced a precision-based…”
Full-page recognition and alignment of historical musical documents
Villarreal, M., Sánchez, J. A., & Parres, D. (2026) · International Journal on Document Analysis and Recognition (IJDAR) · doi:10.1007/s10032-026-00574-w · Cited by 2
“It is important to remark that it was not necessary to apply curriculum learning [41] in order for the model to learn the reading order.”
ALMansour, A. G. M., & Alshomrani, F. (2026) · Scientific Reports · doi:10.1038/s41598-026-48187-x · Cited by 3
“More recently, curriculum learning formulations have been implemented, where the segmentation is progressively optimized through the tumor subregions to ensure stable convergence and balance unequivocal learning and spatial heterogeneity of the classes class 30–32 .”
Nooraiepour, M., Song, Z., Li, W., & Perez, S. (2026) · Energy and AI · doi:10.1016/j.egyai.2026.100757 · Cited by 2
Wang, Y., & Ge, Y. (2026) · Information Systems Research · doi:10.1287/isre.2024.1518 · Cited by 3
A Progressive Learning-to-Learn Knowledge Distillation Method for Out-of-Domain Generalization
Xiang, K., & Fujii, A. (2025) · doi:10.1109/aixse64906.2025.00013 · Cited by 1
“Curriculum Learning proposed by Bengio et al. [6] is a way to train networks by organizing the order in which tasks are learned and incrementally increasing the learning difficulty [7], [8].”
Zhuang, Y., Liu, M., Bai, W., Zhang, Y., Zhang, X., Deng, J., & Ren, F. (2025) · IEEE/CVF International Conference on Computer Vision (ICCV) · doi:10.1109/iccv51701.2025.00440 · Cited by 14
“Inspired by curriculum learning [2], MAR follows a two-stage process.”
Cabld: Contrast-Agnostic Brain Landmark Detection With Consistency-Based Regularization
Salari, S., Harirpoush, A., Rivaz, H., & Xiao, Y. (2025) · IEEE/CVF International Conference on Computer Vision (ICCV) · doi:10.1109/iccv51701.2025.01951 · Cited by 4
“A curriculum learning approach [4, 61] is also employed to mitigate the interferences between the two tasks.”
CaO 2 : Rectifying Inconsistencies in Diffusion-Based Dataset Distillation
Wang, H., Zhao, Z., Wu, J., Shang, Y., Liu, G., & Yan, Y. (2025) · IEEE/CVF International Conference on Computer Vision (ICCV) · doi:10.1109/iccv51701.2025.00449 · Cited by 10
“Therefore, given the input Gaussian noise ε ∼ N ( 0 , I ) , input condition ˆc and the noisy input at time step t , the optimization objective is formulated as: (3) α t is a constant with respect to t , and t is randomly sampled from [1, ˆ T ], where ˆ T ≪ T .”
Diffusion Curriculum: Synthetic-to-Real Data Curriculum via Image-Guided Diffusion
Liang, Y., Bhardwaj, S., & Zhou, T. (2025) · IEEE/CVF International Conference on Computer Vision (ICCV) · doi:10.1109/iccv51701.2025.00166 · Cited by 15
“In contrast, our work aims to design a generative curriculum on a syn-to-real spectrum of data produced by diffusion models, with broader applications in learning from long-tail or low-quality data.”
Wang, M., Li, C., & Vaxman, A. (2025) · IEEE/CVF International Conference on Computer Vision (ICCV) · doi:10.1109/iccv51701.2025.02655 · Cited by 5
“A4.”
Pluskoski, A., Ciganović, I., Jovanović, M., & Vasiljevic, J. (2026) · Electronics · doi:10.3390/electronics15091921 · Cited by 2
Aslam, M. M., Tufail, A., Ding, Y., Silva, L. C. D., Apong, R. A. A. H. M., & Zuhairi, M. F. (2026) · Technologies · doi:10.3390/technologies14050267 · Cited by 1
Singh, K. P., Verma, M., Tyagi, D. K., Vipparthi, S. K., Murala, S., & Abdel-Mottaleb, M. S. (2026) · Knowledge-Based Systems · doi:10.1016/j.knosys.2026.116112 · Cited by 4
Mou, Y., Yuan, G., Song, J., & Li, J. (2026) · IEEE Transactions on Geoscience and Remote Sensing · doi:10.1109/tgrs.2026.3691537 · Cited by 1
“Using static losses (e.g., ERGAS or mse) is typically a ff ected by this issue: during early training, the model prioritizes fitting the dominant low-frequency background to quickly reduce the overall loss, making it di ffi cult to capture high-frequency details [22], [56].”
Hybrid Curriculum Learning for Data-Efficient Lung Nodule Detection with YOLOv11
Luo, Y., Guo, Y., Hooshangnejad, H., et al. (2026) · Diagnostics · doi:10.3390/diagnostics16101441 · Cited by 1
“Curriculum learning (CL) is a training paradigm inspired by human learning, in which training samples are presented according to an estimated difficulty score rather than uniform random shuffling [17].”
Instance-Dependent Early Stopping for Adaptive Data Pruning
Yuan, S., Lin, R., Azian, F., Feng, L., Han, B., Niu, G., Sugiyama, M., & Liu, T. (2026) · IEEE Transactions on Pattern Analysis and Machine Intelligence · doi:10.1109/tpami.2026.3693108 · Cited by 5
“Curriculum Learning [77], [78], [79], [80], [81], [82] is a learning paradigm that aims to improve the efficiency and effectiveness of training by presenting examples in a meaningful order, typically from easy to hard.”
GENRE-CMR: Generalizable Deep Learning for Diverse Multi-domain Cardiac MRI Reconstruction
Hamedani, K. A., Razizadeh, N., Nabavi, S., & Moghaddam, M. E. (2026) · Lecture notes in computer science · doi:10.1007/978-3-032-17734-6_27 · Cited by 2
Alotaibi, N., & BinSaeedan, W. M. (2026) · Drones · doi:10.3390/drones10050394 · Cited by 1
Zhou, Y., Zhang, Z., Pan, J., Lin, Z., & Tu, Z. (2026) · The Visual Computer · doi:10.1007/s00371-026-04522-x · Cited by 1
“This stochastic switching instantiates the curriculum learning framework of Bengio et al. [76], where training begins with richer supervisory signals and progressively transitions to sparser, more abstract conditions.”
Transfer learning reveals large discrepancies between air and land surface temperatures in cities
Zhang, Y., Zhao, L., Chakraborty, T. C., Mazumdar, P., Zhang, K., & Gentine, P. (2026) · Nature Communications · doi:10.1038/s41467-026-73716-7 · Cited by 11
“Curriculum learning prioritizes easier samples at the beginning of training and gradually increases the in fl uence of harder ones as the model improves 80 .”
Face Density as a Proxy for Data Complexity: Quantifying the Hardness of Instance Count
Mohammadi-Seif, A., & Baeza-Yates, R. (2026) · IEEE Conference on Artificial Intelligence (CAI) · doi:10.1109/cai68641.2026.11536291 · Cited by 2
Almohammady, A., Alnanih, R. A., & Alowidi, N. (2026) · Applied Sciences · doi:10.3390/app16115482 · Cited by 1
Xu, T., Chen, C., & Liu, D. (2026) · IEEE Sensors Journal · doi:10.1109/jsen.2026.3698144 · Cited by 1
Surrogate models for rock–fluid interaction: a grid-size-invariant approach
Pinheiro, N. C., Guo, D., Menke, H., Joshi, A. C., Heaney, C. E., Elsheikh, A. H., & Pain, C. C. (2026) · Engineering With Computers · doi:10.1007/s00366-026-02353-x · Cited by 3
“When using curriculum learning, the network is first trained with simpler samples, and then trained with the most complex cases [55, 56].”
Keytar: Practical Keystroke Timing Attacks and Input Reconstruction
Qiu, M., Chuang, L., Kim, D., Qu, H., Chen, T., & Kwong, A. (2026) · IEEE Symposium on Security and Privacy (S&P) · doi:10.1109/sp63933.2026.00106 · Cited by 6
“We introduce curriculum learning, a training paradigm inspired by human cognition, in which a model is sequentially trained on data of increasing difficulty[110, 111].”
A Reference Architecture of Reinforcement Learning Frameworks
Liu, X., & Dávid, I. (2026) · IEEE International Conference on Software Architecture (ICSA) · doi:10.1109/icsa66085.2026.00016 · Cited by 3
“Mava [F16] uses an external checkpoint management library, Orbax [60]. b) Environment Parameter Manager: Handles environment parameters that change over time, e.g., difficulty levels in curriculum learning [61] and parameters in domain randomization methods [62].”
Ramirez, F. M., Clerkin, E. M., Crandall, D., & Smith, L. B. (2026) · Proceedings of the National Academy of Sciences · doi:10.1073/pnas.2536143123 · Cited by 1
Nasr, K. M., Mostafa, S., Maher, A., & Kamel, A. M. (2026) · GPS Solutions · doi:10.1007/s10291-026-02139-0 · Cited by 1
“…followed standard AdamW (Loshchilov 2019) recommendations (learning rate = 𝑠 × 𝑠𝑇 −3 , weight decay = 𝑠 × 𝑠𝑇 −4 , batch size = 𝑠28 ). curriculum thresholds 2009) ( The learning (Bengio et al. 5 s and 10 s) were set based on the distribution of outage durations in the training data,…”
Talukdar, J., & Aspland, T. (2012) · Journal of Curriculum and Teaching · doi:10.5430/jct.v1n1p25 · Cited by 2
“Keller (2010) also asserts in this regard that in South Australia, in spite of sexuality education comprising a „significant component‟ of the health and physical education curriculum doesn‟t validate that it is taught.”
BackgroundFast Optimization of Non-convex Machine Learning Objectives
Νικολάου, Ν. (2012) · doi:10.13140/rg.2.1.3856.6649 · Cited by 1
Yang, Z. (2015) · UKnowledge (University of Kentucky) · Cited by 1
Multi-stage feature learning based object recognition and 3D pose estimation with kinect
Zeng, W., Liang, G., Wang, C., & Wu, X. (2016) · International Conference on Information Science and Technology · doi:10.1109/icist.2016.7483465 · Cited by 1
Robust Large-scale Video Synchronization without Annotations
Wieschollek, P., Freeman, I., & Lensch, H. P. A. (2016) · Cited by 1
Metadata
- Authors
- Bengio, Y., Louradour, J., Collobert, R., & Weston, J.
- Affiliations
- Université de Montréal (CA)
- NEC Laboratories America, Princeton, NJ
- ORCID
- Bengio, Yoshua
- Venue
- Proceedings of the 26th Annual International Conference on Machine Learning
- Publisher
- ACM
- Published
- 2009-06-14
- Pages
- 41-48
- OpenAlex
- W2296073425
- Semantic Scholar
- 8de174ab5419b9d3127695405efd079808e956e8
- Cited by
- 5,366
- References
- 40
- Citations by year
- 2026: 526 · 2025: 636 · 2024: 551 · 2023: 591 · 2022: 536 · 2021: 711 · 2020: 565 · 2019: 442 · 2018: 293 · 2017: 198 · 2016: 103 · 2015: 63 · 2014: 41 · 2013: 40 · 2012: 33
- FWCI
- 16.17
- Type
- conference paper
- Language
- en
- Open access
- closed
- Locations
- International Conference on Machine Learning (publishedVersion)
- Queen Mary Research Online (Queen Mary University of London) (submittedVersion)
- Funders
- Canadian Institute for Advanced Research
- Topics
- Physical Sciences › Computer Science › Artificial Intelligence › Stochastic Gradient Optimization Techniques
- Physical Sciences › Computer Science › Artificial Intelligence › Reinforcement Learning in Robotics
- Physical Sciences › Computer Science › Artificial Intelligence › Neural Networks and Applications
- Keywords
- curriculum learning, training order, non-convex optimization, deep neural network, stochastic neural networks, deterministic neural networks, generalization, convergence rate, numerical continuation, global optimization
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.