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 4,601–4,700 of 5,239
Loss Adaptive Curriculum Learning for Ground-Based Cloud Detection
Qi, T., Hu, Y., & Wang, J. (2025) · Remote Sensing · doi:10.3390/rs17132262
Feldmeier, P., & Fraser, G. (2025) · ACM Transactions on Evolutionary Learning and Optimization · doi:10.1145/3747189 · Cited by 1
“The target selection procedure is aligned with Curriculum Learning [Bengio et al., 2009], in which networks are trained on increasingly more challenging tasks, allowing them to solve complex tasks by improving prior solutions on simpler tasks.”
MCEF: Multi-curriculum Enhancement Framework for Social Image-Text Sentiment Classification
Huang, H., Pan, Y., & Fan, Y. (2025) · Communications in computer and information science · doi:10.1007/978-981-96-6951-6_9
Jhang, S., Yoo, D., Kwon, J., et al. (2025) · The Journal of Korean Institute of Information Technology · doi:10.14801/jkiit.2025.23.6.11
CDL-UNet: Curriculum-Driven Lightweight 3D U-Net for Abdominal Organ Segmentation
Wang, Z., Lin, X., Shi, J., et al. (2025) · Lecture notes in computer science · doi:10.1007/978-3-031-96202-8_8
Search or split: policy gradient with adaptive policy space
Tedeschi, G., Papini, M., Metelli, A. M., & Restelli, M. (2025) · Machine Learning · doi:10.1007/s10994-025-06820-2 · Cited by 1
“In its original formulation, CL (Bengio et al., 2009) proposed “training from easier to harder data", i.e., training the machine learning model with subsets of easy examples from the dataset, then gradually increasing the difficulty until the whole training dataset is considered.”
Bias-Aware Curriculum Sampling For Fair Ranking
Seyedsalehi, S., Le, H. S., Zihayat, M., & Bagheri, E. (2025) · International ACM SIGIR Conference on Research and Development in Information Retrieval (SIGIR) · doi:10.1145/3726302.3730171 · Cited by 1
“Building on curriculum learning principles [6, 18, 31, 36], we hypothesize that introducing less (or high, depending on curriculum direction) biased samples earlier in training may shape the model’s understanding of bias and possibly mitigates bias.”
Song, M., Liu, T., Zhang, W., et al. (2025) · International ACM SIGIR Conference on Research and Development in Information Retrieval (SIGIR) · doi:10.1145/3726302.3730180
“For the second problem, we borrow the idea of curriculum learning [2] that argues neural networks can be trained better when training data is organized in a meaningful order instead of randomly shuffling, which has manifested its advantages in many challenged scenarios and tasks such as few-shot…”
Zhou, B., Wang, C., Wu, X., Liao, C., Wang, P., & Wang, H. (2025) · Lecture notes in computer science · doi:10.1007/978-981-96-9794-6_43
Curriculum Based Reinforcement Learning for 3D Control of Magnetic Microrobot Swarms
Park, M., Sitti, M., & Yoon, J. (2025) · International Conference on Manipulation, Automation and Robotics at Small Scales (MARSS) · doi:10.1109/marss65887.2025.11072792
Deng, J., Chen, R., Cheng, J.-K., & Ma, J. (2025) · Communications in computer and information science · doi:10.1007/978-981-96-9994-0_36 · Cited by 1
Fed-RLCL: Federated Reinforcement Learning with Curriculum Learning for Adaptive Content Sequencing
Rehman, I. U., Ali, Z., & Jan, Z. (2025) · International Conference on Innovation in Artificial Intelligence and Internet of Things (AIIT) · doi:10.1109/aiit63112.2025.11082825 · Cited by 2
“[14] introduced curriculum learning in machine learning, demonstrating its potential to enhance model performance on complex tasks.”
Learning to Assemble with Alternative Plans
Wang, Z., Liu, W., Wang, J., Vallat, G., Shi, F., Parascho, S., & Kamgarpour, M. (2025) · ACM Transactions on Graphics · doi:10.1145/3730824 · Cited by 9
Lin, Y., Ye, C., Zhang, H., Wang, K., Xu, L., Liu, S., & Zhang, Z. (2025) · Lecture notes in computer science · doi:10.1007/978-981-96-9894-3_5 · Cited by 2
Wang, L., & Huang, Y. (2025) · Lecture notes in computer science · doi:10.1007/978-981-96-9894-3_14
Image Transformation Sequence Retrieval with General Reinforcement Learning
Mas-Candela, E., Ríos-Vila, A., & Calvo-Zaragoza, J. (2025) · Lecture notes in computer science · doi:10.1007/978-3-031-99565-1_26
Silent Speech Recognition with Wearable MagnetometersPreprint
Dash, D., Kittle, E., Gerrard, I., Csáky, R., Gonzalez, G., Taylor, D., Llinas, J. P., Labanowski, D., Deka, N., & Yun, R. (2025) · bioRxiv (Cold Spring Harbor Laboratory) · doi:10.1101/2025.08.04.668236 · Cited by 2
“…a deep learning framework that integrates depthwise separable convolutions (Kaiser et al., 2017), squeeze-excitation (SE) blocks (Hu et al., 2018), multi-scale attention mechanisms (Chen & Shi, 2021), and curriculum learning-driven domain adaptation (Bengio et al., 2009; Dash et al., 2023).”
Autonomous Maneuver Decision-Making for Close Air Combat Based on LSTM-DDPG Algorithm
Yang, S.-J., Li, H., Hong, Y., et al. (2025) · Chinese Control and Decision Conference (CCDC) · doi:10.1109/ccdc65474.2025.11090447
“To address these issues, the proposed algorithm incorporates a progressive training framework based on curriculum learning (CL) [23], as shown in fig.”
Zhang, B., Wang, H., Guo, J., et al. (2025) · Chinese Control and Decision Conference (CCDC) · doi:10.1109/ccdc65474.2025.11090906
“To accelerate the training process, a multi-stage training strategy is utilized based on the concept of curriculum learning [17] which has be successfully applied in SARL [18].”
Detecting Large Language Model Generated Korean Essays Using Curriculum LearningPreprint
Jung, Y., & Oh, H. (2025) · SSRN Electronic Journal · doi:10.2139/ssrn.5380275
Frickenstein, L., Thoma, M., Mori, P., Sampath, S. B., Fasfous, N., Vemparala, M. R., Frickenstein, A., Unger, C., Passerone, C., & Stechele, W. (2025) · IEEE Intelligent Vehicles Symposium · doi:10.1109/iv64158.2025.11097516
“Usually, entire training samples are presented in an easy to hard order, inspired by the concept of human learning, defined through manually [9, 11] or automatically [12, 23] designed curricula.”
Jang, B., Kim, M. K., Kim, G. I., Lee, Y. H., Park, C.-R., Ahn, J.-P., & Kim, J. W. (2025) · SSRN Electronic Journal · doi:10.2139/ssrn.5352125
Adaptive Select Loss Strategy for Semantic Segmentation of Agricultural Crop Images
Florea, C., Florea, L., & Ivanovici, M. (2025) · IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing · doi:10.1109/jstars.2025.3589635
“We recall also an important result [31], from neural networks, that is obviously inherited by deep networks [32]; this result states that in the iterative learning process, in the early stages, the clear and simple examples are addressed first (as they produce the largest changes in the derivative),…”
Teaching Recombinable Motifs Through Simple Examples
Ham, H., Zhao, B., Griffiths, T. L., & Vélez, N. (2025) · Cognitive Science · doi:10.1111/cogs.70103
“…of training can speed learning in a variety of domains, such as providing shorter sentences to teach grammars (Elman, 1993), simpler shapes to aid visual classification (Bengio, Louradour, Collobert, & Weston, 2009), or easier motor tasks to aid skill learning (Karpathy & Van de Panne, 2012).”
Chang, X., Wang, X., Huang, X., et al. (2025) · Applied Sciences · doi:10.3390/app15158648
Denisov, D., Feldman, D., Dolev, S., & Segal, M. (2025) · International Conference on Computing and Artificial Intelligence (ICCAI) · doi:10.1109/iccai66501.2025.00047
“Another similar initialization method is curriculum learning [7], where the data is gradually revealed to the network.”
Ultrasound Sensor Data Enhancement via Hybrid Deep Learning Approaches
Marill, L., Bentaleb, A., Maussang, F., et al. (2025) · doi:10.1109/oceans58557.2025.11104570
Curriculum learning with a hierarchical cellular encoding
Vassallo, G., Nisioti, E., Pedersen, J. W., et al. (2025) · Genetic and Evolutionary Computation Conference Companion (GECCO Companion) · doi:10.1145/3712255.3726672
“In this work, we focus on the question of how a genotype to phenotype (GP) map can facilitate evolving solutions of increasing complexity, a setting that has been termed as curriculum learning [1].”
Reinforcement Learning for Autonomous Surface Painting with a Robotic ArmPreprint
Berrocal, E., Merino, I., Montaño, A., & Sierra, B. (2025) · SSRN Electronic Journal · doi:10.2139/ssrn.5396199 · Cited by 1
Application of Language Learning Methodologies in Portuguese Sign Language Translation
Seabra, B., Oliveira, A. I., Sousa, J., & Ferreira, J. E. V. (2025) · International Young Engineers Forum (YEF-ECE) · doi:10.1109/yef-ece66503.2025.11117339
“This progressive way of training the model, called curriculum learning [33], is strongly inspired on the way children start to learn words and, as they develop their vocabulary, they create simple sentences whose complexity increases throughout their language development, as suggested in [12].”
Adaptive Data and Task Joint Scheduling for Multi-Task Learning
Liu, Z., Chai, H., Li, C., Wang, L., & Liao, Q. (2025) · IEEE International Conference on Data Engineering (ICDE) · doi:10.1109/icde65448.2025.00155
“Curriculum learning (CL) is a training paradigm that suggests the order of data can affect the performance of the model [34], [38].”
A Data Pruning Method with Feature Distillation for Improved Computational Efficiency
Soricelli, M., Thompson, R., Chang, Y., & Hixenbaugh, C. (2025) · IEEE International Conference on AI and Data Analytics (ICAD) · doi:10.1109/icad65464.2025.11114026
“Our approach draws inspiration from the literature that focuses on data-centric methods to improve training efficiency [1], [4], [9], [15], [17], [19], [21]–[26].”
Transfer Learning for a Class of Cascade Dynamical Systems
Rabiei, S., Mishra, S., & Paternain, S. (2025) · Proceedings of the ... American Control Conference/Proceedings of the American Control Conference · doi:10.23919/acc63710.2025.11107987 · Cited by 1
“Curriculum learning trains agents progressively with increasing complexity to aid generalization [25].”
Reasoning Distillation and Structural Alignment for Improved Code GenerationPreprint
Jalilifard, A., Raimundo, M. M., & Rocha, A. (2025) · SSRN Electronic Journal · doi:10.2139/ssrn.5403116 · Cited by 1
BOOST: Batch Ordering for Training Deep Forecasting Models
Anaraki, J. R., & Guo, L. (2025) · IEEE Annual International Computer Software and Applications Conference (COMPSAC) · doi:10.1109/compsac65507.2025.00043
“CL was introduced by Benjio and colleagues in 2009 [14], proposing that the learner’s performance can be affected by the order in which samples are presented, inspired by the way humans and animals learn [15].”
Ewers, J., Gibbs, J., & Anderson, D. (2025) · International Conference on Information Fusion (FUSION) · doi:10.23919/fusion65864.2025.11123969
“The ability to change MTT objects within a Gymnasium environment also enables future curriculum learning [14] approaches where agents are trained in environments of increasing complexity to improve training times.”
Quercia, A., Nader, F., Morrison, A., Scharr, H., & Assent, I. (2025) · Knowledge and Information Systems · doi:10.1007/s10115-025-02563-7
“…up training by online sampling proportional to gradient norms via an unbiased estimator [13–15] (or approximations [16–21]), biasing training to hard samples [22–30], presenting samples through a curriculum [31–34], or adaptively subsetting training data to approximate the full gradient [35–38].”
Top-k Bottom All but σ Loss Strategy for Medical Image Segmentation
Florea, C., Florea, L., & Vertan, C. (2025) · Diagnostics · doi:10.3390/diagnostics15172189
“Two contemporary learning strategies, namely curriculum learning [22] and self-paced learning [23], structure the training process in several iterative phases, moving from simple examples to more challenging ones based on individual loss levels.”
BayesIntuit: A Neural Framework for Intuition-Based Reasoning
Bornacelly, M. (2025) · Communications in computer and information science · doi:10.1007/978-3-031-98235-4_9 · Cited by 1
UAV-Enabled Dynamic Data Collection and Energy Replenishment in Large-Scale IoT Networks
Wu, P., Liu, X., Zhu, H. B., et al. (2025) · Proceedings/Proceedings - International Conference on Computer Communications and Networks · doi:10.1109/icccn65249.2025.11133979
“In its original formulation, Bengio et al. [11] introduced curriculum learning as a method to progressively increase the complexity of data samples used during training, aiming to enhance model efficiency and performance.”
Darabi, N., Jayasuriya, D., Naik, D., et al. (2025) · Proceedings - IEEE International Conference on Robotics and Automation/Proceedings · doi:10.1109/icra55743.2025.11128558
“CT, introduced by Bengio et al. [34], mimics the human learning process: starting with easier concepts and gradually progressing to more complex ones to ensure that the model is neither overwhelmed by challenging examples nor constrained by overly simple ones.”
A Data-Efficient Progressive Learning Framework for Robot Scooping Task
Wang, S., Wang, E., Huang, B., Zhang, C., Wang, W., & Zheng, Y. (2025) · Proceedings - IEEE International Conference on Robotics and Automation/Proceedings · doi:10.1109/icra55743.2025.11127485
“To address this, curriculum learning [25] has been proposed as an effective technique to enhance the performance and generalization capabilities of RL models [26].”
Öğretmenlerin Öğretim Programı Farkındalıklarının Öğretim Programı Okuryazarlıklarına Katkısı
Ocak, G., Yılmaz, M., & KARAKUYU, A. (2025) · Abant İzzet Baysal Üniversitesi Eğitim Fakültesi Dergisi · doi:10.17240/aibuefd.2025..-1557815
Hu, T., Zhang, Z., Zhu, C., et al. (2025) · Proceedings - IEEE International Conference on Robotics and Automation/Proceedings · doi:10.1109/icra55743.2025.11128009
“We adopt a curriculum learning approach [23] by dividing the training process into three phases to accelerate the convergence of the policy, as shown in Fig.”
Boosting Hierarchical Reinforcement Learning with Meta-Learning for Complex Task Adaptation
Khajooeinejad, A., Masoumi, F. S., & Chapariniya, M. (2025) · IEEE International Conference on Computing and Machine Intelligence (ICMI) · doi:10.1109/icmi65310.2025.11141230 · Cited by 3
“Curriculum learning complements this by sequencing tasks with increasing difficulty, helping agents build foundational skills before tackling complex problems [13], [14].”
Yee, N. J., Taseh, A., Ghandour, S., et al. (2025) · medRxiv · doi:10.1101/2025.09.04.25334786
Tiered Spatio-Temporal Difficulty: Curriculum Scheduler for Multi-Sensor Traffic Flow Prediction
Zhang, Z., Wang, H., Fan, Z., Jiang, R., Yuan, W., Song, X., & Shibasaki, R. (2025) · IEEE Transactions on Mobile Computing · doi:10.1109/tmc.2025.3608620 · Cited by 3
“Definition 3 ST Model Training with Curriculum Learning [11], [49]: Let z be a sample in training set D , a curriculum is to generate a sequence of training criteria in every model training step: where W t ( z ) denotes the weight for the sample z in t stage.”
InfluentialReview of research progress on self-training for object detection in unmanned system perception
Yu, P., Zhang, X., Sun, H., & Zheng, N. (2025) · Scientia Sinica Informationis · doi:10.1360/ssi-2025-0095
SSCL: Semi-Supervised Comprehensive Learning for Nighttime Semantic Segmentation
Liu, X., Hao, L., Wang, Y., & Wang, J. (2025) · IEEE Transactions on Intelligent Transportation Systems · doi:10.1109/tits.2025.3604412 · Cited by 1
“To further utilize the unlabeled data, we follow the idea of curriculum learning [32], where the model gradually learns from simple to complex examples, and take a further step to adapt it to a dual-mode interactive enhanced version.”
PatLabor: Pareto Optimization of Timing-Driven Routing Trees
Chen, Z., Yao, H., & Yin, X. (2025) · ACM/IEEE Design Automation Conference (DAC) · doi:10.1109/dac63849.2025.11133098
“Motivated by the philosophy of curriculum learning [30], we choose different values of α ( n ) for different degrees of n .”
Jiang, Y., & Chiang, H. (2025) · Artificial Intelligence for Engineering · doi:10.1049/aie2.12003
Mori, M., & Kogawa, Y. (2025) · Lecture notes in computer science · doi:10.1007/978-3-032-05060-1_14
FixGrower: An efficient and robust curriculum for shaping fixation behavior in rodentsPreprint
Breda, J. R., Charlton, J. A., Willock, J. M., Kopec, C. D., & Brody, C. D. (2025) · bioRxiv (Cold Spring Harbor Laboratory) · doi:10.1101/2025.09.12.675850
Trajectory Tracking Control of Autonomous Surface Ships Based on TD3 and Curriculum Learning
Liu, J., Zhou, K., Li, S., et al. (2024) · IEEE International Conference on Underwater System Technology: Theory and Applications (USYS) · doi:10.1109/usys62456.2024.11116221
The Role of Feature Vector Scale in the Adversarial Vulnerability of Convolutional Neural Networks
Park, H.-C., & Lee, S.-W. (2025) · Mathematics · doi:10.3390/math13183026
FastDTI: A 3D Scale-Arbitrary Super-Resolution Autoencoder Residual Dense Network for DTI
Jain, A., Ramanarayanan, S., Ram, K., et al. (2025) · Lecture notes in computer science · doi:10.1007/978-3-032-05573-6_13
The curriculum effect in visual learning: The role of readout dimensionality
Volk, C., Pack, C. C., & Bakhtiari, S. (2026) · PLoS Computational Biology · doi:10.1371/journal.pcbi.1014553 · Cited by 1
“The first claims that easy-to-hard, or simple-to-complex, ordering of training data can help models learn faster and generalize better [33,50–53].”
Sun, K., Han, Y., Zhao, Z., et al. (2025) · 2022 IEEE 18th International Conference on Automation Science and Engineering (CASE) · doi:10.1109/case58245.2025.11163960
“Curriculum learning, inspired by human learning, optimizes machine learning models by progressively introducing tasks of increasing complexity [11].”
Optimizing AI-Driven Production in Industry 4.0: A Morphological Box and Taxonomy Approach
Koval, L., Akmal, M. U., Asif, S., Mathias, S. G., Knollmeyer, S., & Großmann, D. (2025) · International Conference on Computer Technology Applications (ICCTA) · doi:10.1109/iccta65425.2025.11166106
Curriculum Learning Guided Musculoskeletal Grasping Task
Huang, L., Yang, Y., Zhao, J., et al. (2025) · IEEE International Conference on Cyber Technology in Automation, Control, and Intelligent Systems · doi:10.1109/cyber67662.2025.11168323
“Curriculum Learning [1] builds upon the principles of Reinforcement Learning (RL) [2], where agents learn to make decisions by interacting with an environment and adjusting their policy based on feedback.”
Totmina, E. V., Bondarenko, I., & Seredkin, A. V. (2025) · Modeling and Analysis of Information Systems · doi:10.18255/1818-1015-2025-3-282-297
Early lifespan prediction in Caenorhabditis elegans via contrastive learning and channel attention
Jin, M., Chen, W., & Pan, Y. F. (2025) · Journal of Bioinformatics and Computational Biology · doi:10.1142/s0219720025500167
Liu, P., Sun, F., Zhao, L., Fang, W., & Huang, F. (2025) · The Computer Journal · doi:10.1093/comjnl/bxaf110
Communication-Driven Multi-Agent Path Planning Based on Improved Attention Mechanism
Sheng, Y.-H., & Yin, H. (2025) · Lecture notes in mechanical engineering · doi:10.1007/978-3-031-95455-9_9
Sun, Z., Hua, Z., Li, H., et al. (2025) · Signal Image and Video Processing · doi:10.1007/s11760-025-04795-0
“Amachinelearningparadigmmimickinghumanlearning— curriculum learning [9, 10]—introduces easy examples first, gradually adding harder ones as learning progresses.”
Exploring Curriculum Learning for Languages: Lessons from Regular Language Tasks
Toborek, V., Seiffarth, F., Müller, S., Horváth, T., & Bauckhage, C. (2025) · Lecture notes in computer science · doi:10.1007/978-3-032-05461-6_37
Shen, C.-W., Jiang, J., & Hsieh, H.-P. (2025) · IEEE Transactions on Big Data · doi:10.1109/tbdata.2025.3618482 · Cited by 1
“Scheduled sampling [5], [72] addresses exposure bias issues encountered by models with an encoder-decoder architecture, while curriculum learning [8], [73] aids the model in finding a better local optimum.”
Sun, Z., Zhang, W., Zhou, Y., Geng, S., Zhang, D., Wang, J., Liu, B., Fu, Z., Zheng, L., Jiang, C., et al. (2025) · PLOS Digital Health · doi:10.1371/journal.pdig.0001037 · Cited by 3
“Curriculum Learning (CL) [29] is a training strategy that encourages models to learn progressively from easy to difficult examples.”
InfluentialReinforcement Learning Identifies Age-Related Balance Strategy Shifts
Wang, H., Kövecses, J., & Durandau, G. (2025) · IEEE Transactions on Neural Systems and Rehabilitation Engineering · doi:10.1109/tnsre.2025.3619868 · Cited by 2
“Similarly, machine learning benefits from a similar training strategy by starting with easier subtasks and gradually increasing the level of di ffi culty [42], [43].”
Dual-stage framework with soft-label distillation and spatial prompting for image-text retrieval
Jin, R., Li, Z., Deng, F. Y., Zhang, Y., Luo, M., Jin, T., Hou, T., Du, C., Gu, X., & Yuan, J. (2025) · PLoS ONE · doi:10.1371/journal.pone.0333084
“Furthermore, we incorporate a curriculum [6] learning strategy to progressively refine the training process, enabling the model to learn more effectively and ultimately achieve superior performance.”
OPTIMIZATION OF TRAINING SAMPLE USING RANDOM POINT PROCESSES
Lutsyk, O., Rusyn, B., & Kosarevych, R. (2025) · Information and communication technologies electronic engineering · doi:10.23939/ictee2025.02.109
Enhancing generalization in autonomous driving through track-agnostic reinforcement learning
Authier-Carcelen, J.-B., & Zadourian, R. (2025) · Neural Computing and Applications · doi:10.1007/s00521-025-11597-5
“Hierarchical learning and knowledge transfer are foundational to curriculum learning, where models are trained on simpler tasks before progressing to more complex ones [41].”
Sliding RandAugment: Dynamic Dual-Window Data Augmentation for Efficient Traffic Sign Recognition
Drdák, P., & Lehoczký, P. (2025) · International Symposium ELMAR (ELMAR) · doi:10.1109/elmar66948.2025.11194024
“Applied early in training, they help the model learn basic traffic sign features [8], [9].”
Fan, S., Zhou, L., Sun, J., et al. (2025) · WRC Symposium on Advanced Robotics and Automation (WRC SARA) · doi:10.1109/wrcsara68202.2025.11195032
“Curriculum learning (CL), first proposed by Bengio et al. in 2009[14], is founded on the idea of training a model on simpler examples before gradually exposing it to more complex ones, thereby accelerating convergence and improving generalization.”
RealCustom++: Representing Images as Real Textual Word for Real-Time Customization
Mao, Z., Huang, M., Ding, F., Liu, M., He, Q., & Zhang, Y. (2025) · IEEE Transactions on Pattern Analysis and Machine Intelligence · doi:10.1109/tpami.2025.3623025 · Cited by 33
“Curriculum learning [46], [47], [48], [49] is a training strategy that organizes data from “easy” to “hard” to mimic human learning, aiming to improve model performance and accelerate convergence.”
OOB-CM: Enhancing OOB Estimate for Data Valuation via Curriculum Learning and Multi-round Voting
Jiao, Y., Sun, G., Wang, Z., et al. (2025) · Lecture notes in computer science · doi:10.1007/978-981-95-1346-8_12
Merze, A., & Çelebi, F. V. (2025) · Engineering Science and Technology, an International Journal · doi:10.1016/j.jestch.2025.102205 · Cited by 2
Learning Whole-Body Motion Control Through Instruction Learning and Human Motion Data
Xu, Z., Chen, K., Ye, L., & Xing, B. (2025) · Lecture notes in computer science · doi:10.1007/978-981-95-2098-5_59
Tang, X., Xiang, Y., Yu, Z., & Huang, Y. (2025) · IEEE International Conference on Multimedia and Expo (ICME) · doi:10.1109/icme59968.2025.11209332
“To account for varying degrees of modality missingness, we design a Curriculum Learning-based (CL) [13] training strategy that gradually increases the learning difficulty by simulating missing modalities from simple to complex scenarios, enabling the model to adapt to diverse and challenging…”
DLLM: Enhancing Open-World Object Detection with Dynamic Learning and Large Models
Huang, Y., Xi, X., & Luo, R. (2025) · IEEE International Conference on Multimedia and Expo (ICME) · doi:10.1109/icme59968.2025.11209937
D2AD: Diffusion Distillation for Unsupervised Image Anomaly Detection
Shao, Y., Ni, Z., & Liu, Q. (2025) · IEEE International Conference on Multimedia and Expo (ICME) · doi:10.1109/icme59968.2025.11209101
“Inspired by the concept of curriculum learning [24], D2AD is designed in two stages: the Normality Pretraining Stage (Sec.”
Mihhailov, K., Hou, M., & Baraka, K. (2025) · IEEE RO-MAN · doi:10.1109/ro-man63969.2025.11217772
“Curriculum learning [31], proposes structuring training examples from simple to complex to enhance learning efficiency.”
Teachable Social Robots: Managing Expectations in Highly Anthropomorphic Designs
Majumder, T., Ashok, A., Rosén, J., Sevinc, A., & Berns, K. (2025) · IEEE RO-MAN · doi:10.1109/ro-man63969.2025.11217750
“Curriculum Learning (CL) [11] supports robot learning by breaking tasks into increasingly complex subtasks, enabling social robots to deliver personalized skills [6].”
Goal-aware Sparse GNN for RL-based Generalized Planning: Towards Practical Drone Mission Planning
Jeon, S., Shin, J., Kim, G.-T., Cho, Y., & Kim, S. (2025) · International Journal of Control Automation and Systems · doi:10.1007/s12555-025-0543-2 · Cited by 2
“Curriculum learning progressively increases task difficulty during training, enabling agents to first master simpler tasks and subsequently build upon these acquired skills[11].”
Ahmed, N. S., & Nower, N. (2025) · Procedia Computer Science · doi:10.1016/j.procs.2025.09.457 · Cited by 1
Curriculum Guided Personalized Subgraph Federated Learning
Kang, M., & Park, H. (2025) · ACM International Conference on Information and Knowledge Management (CIKM) · doi:10.1145/3746252.3761128
Hsu, B.-Y., Lu, C.-H., & Shen, C.-Y. (2025) · ACM International Conference on Information and Knowledge Management (CIKM) · doi:10.1145/3746252.3761139
From Coarse to Fine, Easy to Hard: Dual-Progressive 3D Hand Mesh Reconstruction
Lu, Y., Zuo, Z., & Sang, N. (2025) · Lecture notes in computer science · doi:10.1007/978-981-95-4395-3_9
FixCLR: Negative-Class Contrastive Learning for Semi-Supervised Domain Generalization
Son, H. M., Rezaei, S., & Liu, X. (2025) · IEEE International Conference on Multimedia Information Processing and Retrieval (MIPR) · doi:10.1109/mipr67560.2025.00027
“As it is standard to perform leave-one-domain-out evaluation, pretrained models may have an advantage in such cases as they have been exposed to similar domain information during pretraining.”
Huh, Y., & Seo, Y. (2025) · ACM International Conference on AI in Finance (ICAIF) · doi:10.1145/3768292.3770384
Brunner, A., Kotsis, G., & Khalil, I. (2025) · Lecture notes in business information processing · doi:10.1007/978-3-032-08603-7_8
Skaf, W., Baratchi, M., & Hoos, H. H. (2025) · Machine Learning · doi:10.1007/s10994-025-06895-x · Cited by 1
“…remain constant throughout the training or can be changed following a given schedule, a curriculum inspired by the concept of curriculum learning (Bengio et al, 2009), which offers more flexibility and capacity to handle different cases of teacher and student models with different performance…”
InfluentialCurriculum-Meta learning for unbiased Multimodal Relation Extraction
Su, X., Jin, Y., Wang, Z., Liu, C., Wang, A., & Chen, P. (2026) · Journal of Visual Communication and Image Representation · doi:10.1016/j.jvcir.2026.104834
Adaptive Sample Weighting with Regime-Aware Meta-Learning Framework for Financial Forecasting
Jang, J., Kim, T., Oh, D., & Park, S.-H. (2025) · ACM International Conference on AI in Finance (ICAIF) · doi:10.1145/3768292.3770374
“Related concepts like curriculum learning proposes training models on simpler examples first before progressing to more difficult ones, mimicking human learning patterns and improving convergence in complex tasks [4].”
Abbas, A., Gaber, M. M., & Abdelsamea, M. M. (2025) · Lecture notes in computer science · doi:10.1007/978-981-95-4381-6_7
Rudroff, T. (2025) · Neurocultural health and wellbeing · doi:10.1007/978-3-032-11596-6_3
Towards Efficient Structured Description Generation for Data Marketplace Offerings
Awan, M., Nadeem, A., Santana, J. R., et al. (2025) · IEEE Conference on Standards for Communications and Networking · doi:10.1109/cscn67557.2025.11230736
“2) Curriculum Learning Approaches: Curriculum learning progressively trains models on increasingly complex or context-scarce examples, improving robustness and schema-generalization capabilities [41].”
HACL: A Hybrid Adaptive Curriculum Learning Framework for Multi-modal Sarcasm Detection
Shen, K., Huang, Y., Wang, W., Xie, S., Zhu, Z., & Liu, W. (2025) · International Joint Conference on Neural Networks (IJCNN) · doi:10.1109/ijcnn64981.2025.11228306
“[20] first formalized CL theory and introduced the Baby Step curriculum [31], sparking subsequent development of sophisticated CL variants including Self-Paced Learning [32], [33], Transfer Teacher frameworks [34], and others [35], [36].”
Enhanced Neural Architecture Search with Multi-Dimensional Curriculum Learning for Defect Detection
Yong, Y., & Ma, L. (2025) · International Joint Conference on Neural Networks (IJCNN) · doi:10.1109/ijcnn64981.2025.11228890
“Curriculum learning is a machine learning training strategy that follows a progressive learning approach: Starting with simpler tasks and gradually introducing more complex ones [9].”
Curriculum Design for Scalable Biologically Plausible Deep Reinforcement Learning
van den Berg, A. R., Roelfsema, P. R., & Bohté, S. M. (2025) · International Joint Conference on Neural Networks (IJCNN) · doi:10.1109/ijcnn64981.2025.11229171
“A curriculum could allow for initial optimisation of an easy function, which can then progressively be refined as it becomes more non-convex through more difficult examples [11].”
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.