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 5,001–5,100 of 5,240
Controle Seguro e Evasão de Colisões em Tráfego Denso de Drones via Aprendizado por Reforço
dos Santos, H. J. F., Barros, I. D. S., Bittencourt, L. F., Kamienski, C., & de Oliveira, F. M. C. (2026) · Anais do Workshop de Computação Urbana (CoUrb) · doi:10.5753/courb.2026.23126
Scheduled Partial-Credit RL for Reliable Code Generation with Small Language Models (WIP)
Sijwali, S. S., & Saha, S. (2026) · ACM SIGPLAN/SIGBED Conference on Languages, Compilers, and Tools for Embedded Systems (LCTES) · doi:10.1145/3814943.3816167
“This is consistent with curriculum learning [1] and motivates reward scheduling as a key design concern for SLM training.”
Tanabe, H. (2026) · SSRN Electronic Journal · doi:10.2139/ssrn.6928679
Li, J., Wu, W., Wang, S., Pan, Z., Lee, K. A., Meng, H. M. L., & Li, H. (2026) · IEEE Transactions on Audio Speech and Language Processing · doi:10.1109/taslpro.2026.3703224 · Cited by 9
“To mitigate this issue, we adopt a curriculum learning strategy [66], [67], [68], gradually transitioning from the ground truth target speech x to the self-enrolled speech ˆ x 1 .”
Aerial Pursuit: Multi-Agent Swarm for Predator–Prey Games
Wu, Z.-Q., Chen, H.-R., Zhao, J., Peng, Y., Srigrarom, S., & Khoo, B. (2026) · 2026 12th International Conference on Electrical Engineering, Control and Robotics (EECR) · doi:10.1109/eecr69522.2026.11549319
“Mainstream approaches include the following: • using PPO, SAC, and other on-policy or off-policy meth-ods with parallel sampling and entropy regularization to mitigate insufficient exploration [1], [2]; • employing curriculum learning to gradually increase task difficulty [3]; and”
Competitive Memory Editing: A Nash Bargaining Approach to MEMIT Interference Prediction
Mohan, M. K., & Sen, A. (2026) · International Conference on Information and Computer Technologies (ICICT) · doi:10.1145/3803291.3803357
“Future work could reduce the quadratic growth of interference through sequential edit scheduling rather than simultaneous application [14], where high-risk edits could be delayed or isolated to reduce cascading effects.”
He, Y., Zhao, S., & Li, S. (2026) · Lecture notes in electrical engineering · doi:10.1007/978-981-95-4897-2_67
Uncertainty-Guided Curriculum Learning for Automated Liver Fibrosis Staging on Heterogeneous MRI
Jin, Y., Zhao, F., Chen, Y., & He, X. (2026) · Lecture notes in computer science · doi:10.1007/978-3-032-16271-7_2
Exploring Potential Knowledge Correlations of Disease Diagnosis with Partial Labels
艾念, Liu, D., & Zhou, P. (2026) · International Conference on Multimedia Retrieval · doi:10.1145/3805622.3810867
“Here, we propose an adaptive curriculum learning strategy [1] to relabel the uncertain parts.”
An Integrated Architecture for Robust End-to-End Radar Multitarget Detection in Jamming Environments
Zhu, H., Liao, K., Wang, Y.-X., et al. (2026) · IEEE Transactions on Radar Systems · doi:10.1109/trs.2026.3703613
“In terms of model training strategy optimization, Bengio et al. [35] first proposed the curriculum learning paradigm, which imitates human cognitive processes, decomposes tasks according to learning di ffi culty, and follows the order of “easy to hard” training.”
星, 克., Liang, D., & Shang, K. (2026) · International Conference on Multimedia Retrieval · doi:10.1145/3805622.3810885 · Cited by 1
“This design aligns multi-objective optimization with cognitive principles by incorporating Curriculum Learning [4] and gradient-based balancing [6], ensuring semantic guidance is applied upon a stabilized visual representation.”
STORM-OPF: Robustness Benchmarking for AC-OPF Methods
Afrin, A., & Ardakanian, O. (2026) · ACM International Conference on Future Energy Systems (e-Energy) · doi:10.1145/3744255.3811719
Progressive Curriculum Learning and Ghost-Aware Supervision for Real-World Face Detection
Guo, Q., Wang, Y., 林, 智., Luo, L., & Luo, Y. (2026) · International Conference on Automatic Face and Gesture Recognition · doi:10.1109/fg67764.2026.11556975
“Curriculum learning (CL) was originally introduced by [2] as a training paradigm that gradually increases task difficulty to stabilize optimization and improve generalization.”
Data‐Efficient Electromagnetic Surrogate Solver Through Dissipative Relaxation Transfer Learning
Nam, S., Park, C. Y., & Jang, M. S. (2026) · Advanced Optical Materials · doi:10.1002/adom.71366
A physics-informed autoencoder for digital twin development in offshore wind turbine installation
Dunton, D., Bhaskaran, S., Zhu, Y., et al. (2026) · Structural and Multidisciplinary Optimization · doi:10.1007/s00158-026-04368-w
“The following sections discuss each loss term introduced during AE training, along with the Curriculum Learning (CL) strategy used to gradually incorporate loss terms of increasing complexity (Bengio et al. 2009).”
A Fast and Accurate Surrogate Model for Clock-Mesh Timing Analysis
Khan, M. H., & Guthaus, M. R. (2026) · ACM Great Lakes Symposium on VLSI (GLSVLSI) · doi:10.1145/3787109.3815287 · Cited by 2
FNeRV: Frequency-Aware Neural Video Representations with Dynamic Spatial Guidance
Xiong, Z., Jiang, Y., Li, L., et al. (2026) · IEEE International Symposium on Circuits and Systems proceedings · doi:10.1109/iscas66217.2026.11562364
“Inspired by curriculum learning [18], the network initially emphasizes low-resolution structures to capture global shapes and overall layout.”
Decompose-learn-revise: A reliability-aware framework for robust time series forecasting
Zhang, X., & Chen, Z. S. (2026) · Computers & Electrical Engineering · doi:10.1016/j.compeleceng.2026.111291
Dual-Path Conditional Chain for CTC-Based Multi-Talker Speech Recognition
Shi, Y., & Han, J. (2026) · IEEE Signal Processing Letters · doi:10.1109/lsp.2026.3705378
“Curriculum learning [37] was adopted: training began with single-speaker and 2-Mix data, and 3-Mix data were introduced when the token error rate dropped below 40%.”
Dynamically robust loss: Unlocking noisy-label learning in named entity recognition
Hua, Z., Huang, J., & Chen, Y. (2026) · Information Processing & Management · doi:10.1016/j.ipm.2026.104991
CSDTab-ID: A Cross-Scale Difficulty-Aware Tabular Transformer for Robust Network Intrusion Detection
C, M. S., Ashok, A., & S, A. G. A. S. G. (2026) · International Conference on Inventive Research in Computing Applications (ICIRCA) · doi:10.1109/icirca69024.2026.11570544
“Curriculum learning shows that progressively approaching difficulty samples which will improve generalization [13].”
Chang, C.-C., & Lin, M.-H. (2026) · Computers & Electrical Engineering · doi:10.1016/j.compeleceng.2026.111341
Adaptive masked autoencoder for anomaly detection
Hu, R., & Chen, Z. (2026) · Neural Computing and Applications · doi:10.1007/s00521-026-12281-y
Wen, T., Xiong, Z., Lian, X., Zhang, J., & Shen, H. (2026) · Chinese Control and Decision Conference (CCDC) · doi:10.1109/ccdc69976.2026.11560804
In-Context Region-Based Drag: Drag Any Region to Any Shape
Sui, J., Hao, T., Gao, B., Niu, L., & Zhai, G. (2026) · Lecture notes in computer science · doi:10.1007/978-3-032-37490-5_26
PINNs for Electromagnetic Wave Propagation
Bulut, N. K. (2026) · Lecture notes in computer science · doi:10.1007/978-3-032-29912-3_12
Scheduled Temporal Loss Weighting for Neural Operators
Coker, O. E., Wang, H., Khan, A., et al. (2026) · Lecture notes in computer science · doi:10.1007/978-3-032-29924-6_15
“Curriculum Learning (CL) improves convergence by transitioning from “easy” to “hard” samples [1], defined by a curriculum metric and a scheduler.”
Yan, K., Ji, T., Xie, F., Liu, C., & Yuan, Z. (2026) · Drones · doi:10.3390/drones10070489
Confidence-Tiered Pseudo-Labeling for Few-Shot Object Detection in Remote Sensing Imagery
Shi, Y.-Q., & Yang, R.-P. (2026) · 2026 6th International Symposium on Computer Technology and Information Science (ISCTIS) · doi:10.1109/isctis70043.2026.11572318
“Curriculum learning [16] trains models by presenting examples in a meaningful order, typically from easy to hard.”
Rollback Is Not Undo: Path-Dependent Failures in LLM-Arbitrated Network Control
Pan, W., & Liu, Z. (2026) · IEEE Conference on Computer Communications (INFOCOM) · doi:10.1109/infocom59046.2026.11571400
“This is a form of path dependence (hysteresis) [5]: the outcome depends on the route taken, not only the endpoint configuration.”
Five Educational-Psychology Lenses for Training and Evaluating AI Models
Wang, P., & PAAS, F. G. W. C. (2026) · Educational Psychology Review · doi:10.1007/s10648-026-10195-8
“This does not replace the engineering role, but it adds a teacher-like design function to model development (Bengio et al., 2009).”
MirrorPPR: Exemplar-Based Portrait Photo Retouching
Liu, Z., Li, Z., Jin, J., et al. (2026) · Lecture notes in computer science · doi:10.1007/978-3-032-37196-6_24
Vision-Language Model Training for Code Generation from Requirements and Model-Based Artifacts
Padubrin, M., Kulzer, A. C., & Guerocak, E. (2026) · SAE technical papers on CD-ROM/SAE technical paper series · doi:10.4271/2026-01-0770
Jiang, N., Hu, J., Sun, B., Sim, T., & Han, J. (2026) · IEEE Symposium on Security and Privacy (S&P) · doi:10.1109/sp63933.2026.00197 · Cited by 1
“To address this issue, we adopt a curriculum training strategy [44].”
Reinforcement Learning on Data-Dependence Graphs for Custom Instruction Identification
Hussein, E., Waschneck, B., & Mayr, C. (2026) · ACM Transactions on Architecture and Code Optimization · doi:10.1145/3828552 · Cited by 1
“Training follows a curriculum over DDG size, I/O bounds, and area factors [7].”
Singh, N. P., Wang, X., Garikipati, A., Ciobanu, M., Mao, Q., & Das, R. (2026) · Frontiers in Artificial Intelligence · doi:10.3389/frai.2026.1851993
“Our data curation process applied gradient-based selection: we eliminated questions the model answered correctly consistently, retaining only items where the model exhibited uncertainty or errors.”
Park, Y. H., Lee, K.-B., & Kim, H. (2026) · Bioengineering · doi:10.3390/bioengineering13070769
“As an efficient alternative, curriculum learning has attracted attention as a strategy that facilitates model convergence by controlling the training order according to data difficulty or information content [10].”
InfluentialSeismic Data Reconstruction Using a Hybrid Dimension Attention Restormer via Progressive Learning
Tang, Y., Wu, H., Zheng, Q., et al. (2026) · Applied Sciences · doi:10.3390/app16136589
Tan, Y. L. J., & Soh, G. S. (2026) · Mechanisms and machine science · doi:10.1007/978-3-032-29469-2_12
Liquid neural network-based multi-UAV path planning in dynamic, obstacle-aware environments
Güzey, A., Çifçi, M. A., Erdoğan, A. Y., & Yildirim, F. (2026) · Discover Artificial Intelligence · doi:10.1007/s44163-026-01684-4
“Action masking restricts the policy output space to feasible actions [24], curriculum learning gradually increases task difficulty [25], and conditional or structured imitation architectures have been used to better represent multimodal expert behavior [26].”
Onan, A., Nasution, A. H., & Eğin, F. (2026) · Discover Computing · doi:10.1007/s10791-026-10329-0
“In parallel, curriculum learning has been shown to stabilize optimization and improve generalization by ordering training examples from easier to harder instances, particularly in low-resource and multilingual scenarios [15–17].”
Dharangaonkar, S., Kataria, A., & Mundhe, S. (2026) · Lecture notes in networks and systems · doi:10.1007/978-3-032-08600-6_20
Multi-granular metro passenger flow forecasting via MMoE and hierarchical multi-source data fusion
Yang, Q., Xu, X., Yu, J., et al. (2026) · Transportmetrica B Transport Dynamics · doi:10.1080/21680566.2026.2696355
Zhang, P., Xiang, D., Huo, C., Ding, K., & Xiang, S. (2026) · Remote Sensing · doi:10.3390/rs18132202
Chen, J. (2026) · International Conference on Computer Supported Cooperative Work in Design (CSCWD) · doi:10.1109/cscwd68734.2026.11581584
“…critical challenges in this collaborative process: (1) the ”noisy teacher” problem [13], by implementing a data sanitization phase to filter out incorrect teacher reasoning; and (2) the inherent difficulty of learning complex reasoning, by introducing a novel Curriculum Learning (CL) strategy [14].”
Nawrocka, A., Nawrocki, M., Chabowski, I., & Kot, A. (2026) · International Carpathian Control Conference · doi:10.1109/iccc71363.2026.11593333
Non-linear control variate in 𝛿𝑓 particle-in-cell methods using symplectic neural networksPreprint
Fournet, V., Pinto, M. C., Franck, E., et al. (2026) · SSRN Electronic Journal · doi:10.2139/ssrn.7070225
Optimization of Robot Grasping Using Reinforcement Learning in Simulation
Belke, M., Phatak, M., Petrović, O., & Brecher, C. (2026) · Springer eBooks · doi:10.1007/978-3-031-91463-8_26
Curriculum-Based Deep Reinforcement Learning for Navigation of Unmanned Ground Vehicles
Zhu, C., Chen, P., Lu, Z., Jin, L., He, X., & Chen, Q. (2026) · International Annual Conference on Complex Systems and Intelligent Science (CSIS-IAC) · doi:10.1109/csis-iac70275.2026.11585023
“Following [19], [20], this paper adopts a progressive curriculum that gradually increases the environment complexity from E 1 to E 3 .”
Multi-Stage Machine Learning and Graph-Based Novel Framework for Predicting Quasar Redshifts
Narkedimilli, S., De, A., V, S. K. T., V, Mishra, O., S, P. K. K. N., & Aswath Babu, H. (2026) · Procedia Computer Science · doi:10.1016/j.procs.2026.06.222
Kim, B., Kim, J., Kim, Y., et al. (2026) · BMC Medical Informatics and Decision Making · doi:10.1186/s12911-026-03692-x
Tripathee, Y., Sultan, N., Iqbal, F., & Prom–on, S. (2026) · International Joint Conference on Computer Science and Software Engineering · doi:10.1109/jcsse68839.2026.11596626 · Cited by 1
“The final uncertainty-weighted loss automatically balances both tasks by learning task-specific uncertainty parameters for detection σ d and captioning σ c : The proposed model applies dataset-specific curriculum learning [17] and gradient surgery [18] to resolve conflicting between two tasks.”
Cai, Y., & Zhao, J. Z. (2026) · Aerospace · doi:10.3390/aerospace13070626
Interpretable Control with Graph-Based Genetic Programming
Nadizar, G. (2026) · Genetic and evolutionary computation · doi:10.1007/978-981-95-6398-2_13
ThaiGEC-Prompt: Few-Shot Prompting with Domain-Specific Examples for Thai Learners
Techadisai, K., & Sirinaovakul, B. (2026) · International Joint Conference on Computer Science and Software Engineering · doi:10.1109/jcsse68839.2026.11596908
“An ablation study in [14] and [15] was conducted to systematically analyze the contribution of individual components within our proposed framework.”
Chen, M., Luo, F., Xie, J., & Ren, Q. (2026) · Bioengineering · doi:10.3390/bioengineering13070794
“The dual-agreement gate selects samples independently corroborated by two structurally dissimilar systems—a deep sequence model and a classical tree ensemble—prioritising gradient quality over gradient quantity, consistent with curriculum learning principles [32].”
Interactive LLM-Assisted Curriculum Learning for Multi-Task Evolutionary Policy Search
Sakallioglu, B., Nadizar, G., & Medvet, E. (2026) · Genetic and Evolutionary Computation Conference (GECCO) · doi:10.1145/3795095.3805152
“Crafting representative, difficulty-calibrated training instances or auxiliary tasks typically requires substantial domain expertise [4, 37].”
InfluentialRAG-GNN: retrieval-augmented graph neural networks for protein interaction network embeddings
Hays, H., & Richardson, W. J. (2026) · Frontiers in Artificial Intelligence · doi:10.3389/frai.2026.1851917 · Cited by 3
“The contrastive loss for node v i with positive document The full loss aggregates over all nodes: The temperature parameter τ controls the concentration of the distribution, with smaller values increasing the penalty for misaligned embeddings.”
A comprehensive review on green AI methods, strategies & measurement frameworks
Bayoumi, R., Nazmy, T., Salem, A. M., & Hindy, H. (2026) · Journal of King Saud University - Computer and Information Sciences · doi:10.1007/s44443-026-00603-y · Cited by 3
“Thetraining scheduler then orchestrates the sequence and pace at which these samples are presented, gradually increasing complexity as the model’s competence improves (Emmanouil et al. 2019; Bengio et al. 2009; Spitkovsky et al. 2010; Haco-hen and Weinshall 2019).”
CuForm: Data-Efficient Formula Generation via Curriculum Learning
Qiu, W., Ma, M., Qiu, H., Wang, J., & Wen, H. (2026) · Lecture notes in computer science · doi:10.1007/978-981-92-2852-2_26
Wu, Y., Yang, Y., Deng, B. Q., & Li, J. J. (2026) · Symmetry · doi:10.3390/sym18071186
Physics-guided multi-task learning for urban flood prediction under drainage sedimentation scenarios
Chu, W., Hu, Y., Jin, C., Zhang, C., Di, S., Li, R., Wang, T., & Li, H. (2026) · Journal of Hydrology · doi:10.1016/j.jhydrol.2026.136036 · Cited by 2
Suzuki, I., & Hartono, P. (2026) · IEEE International Conference on Human-Machine Systems (ICHMS) · doi:10.1109/ichms69701.2026.11602341
“Many attempts have been made to study how humans interact with AI in HITL-ML [1], [6], [5].”
HD-CLIP: Hierarchical Dynamic Prompting and Decoupled Learning for Zero-Shot Anomaly Detection
Jiang, J., Chang, Y., Yin, H., Wei, S., & Xu, X. (2026) · IEEE Transactions on Instrumentation and Measurement · doi:10.1109/tim.2026.3711336 · Cited by 1
“We adopt a two-stage regimen following curriculum learning principles [18], [28].”
Wang, H., Zhang, Y., Karrenbach, M., Ding, Y., & He, B. (2026) · Nature Communications · doi:10.1038/s41467-026-75435-5 · Cited by 4
“…progressively increasing inclusion threshold implements a curriculum-style progression inspired by curriculum and self-paced learning paradigms 64,65 , in which optimization begins with highly reliable samples and gradually incorporates more challenging ones, improving convergence stability and…”
Student-Teacher Error-Aware Displacement for Semi-supervised Medical Image Segmentation
Wei, Y., Ma, L., & Zhang, H. (2026) · Lecture notes in computer science · doi:10.1007/978-981-92-3420-2_22
Dynamic Curriculum Learning over Difficulty Heterogeneity for Text-to-SQL
Zhang, X., Tao, C., Xing, S., Chen, Z., & Ma, S. (2026) · Lecture notes in computer science · doi:10.1007/978-981-92-2759-4_12
Spiking Neural Networks for Real-Time Strategy: A Curriculum-Trained SNN Agent for MicroRTS
Liu, C. (2026) · International Conference on Neuromorphic Systems (ICONS) · doi:10.1145/3822454.3822461
“95) across three curriculum stages [3]: Stage 1 teaches resource harvesting against a passive (no-op) opponent.”
CoGrad: Profiling Competence Boundaries for Efficient Mathematical Alignment in LLMs
Zhang, Y., Wu, W., Zhao, Y., Liu, X., & Liu, S. (2026) · Lecture notes in computer science · doi:10.1007/978-981-92-3417-2_21
Mohammad, F., Kakar, J. K., NDONG, D. R. B. B., Chas, M., & Ryu, D. (2026) · IEEE International Conference on Software Testing, Verification and Validation Workshops (ICSTW) · doi:10.1109/icstw72326.2026.00035
“Training follows a three-phase curriculum [26]: foundation (easy, clear-metric examples, epochs 1–3), development (nuanced, conflicting-metric cases, epochs 4–6), and mastery (hard, adversarial, and Teacher-disagreement cases, epochs 7–10).”
Chen, C., Shi, J., Zhang, D., Yao, L., Liu, T., Deng, C., Tian, X., Ge, H., Luo, G., & Liu, Y. (2026) · IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing · doi:10.1109/jstars.2026.3714475
“Furthermore, to alleviate the training difficulties caused by the large action space and scarce effective feedback, we introduce a CL strategy [34], [35] to gradually increase the task difficulty, thereby helping the agent learn effective localization policies more easily in the early stages.”
Uniyal, D., Bashar, M. A., & Nayak, R. (2026) · Intelligent Systems with Applications · doi:10.1016/j.iswa.2026.200706
MGL-LMM: Generative OCR with Semantic Inpainting for Degraded Mongolian Lead-Type Printed Documents
Luo, Z., Lu, M., Liu, N., Shi, B., 仁庆道尔吉, & Wu, N. (2026) · Lecture notes in computer science · doi:10.1007/978-981-92-3520-9_46
A Cost-Sensitive Three-Way Routing Framework for Edge-Cloud LLMs
Zhao, P., Huang, D., Zhang, Q., et al. (2026) · Communications in computer and information science · doi:10.1007/978-981-92-3557-5_17
Boost-BAAT: Hardness-Aware Sample Selection for Clean-Label Backdoor Attacks
Jiang, S., & Xia, Z. (2026) · Lecture notes in computer science · doi:10.1007/978-981-92-3423-3_14
Patel, R., & Salaymeh, A. (2026) · Drones · doi:10.3390/drones10070548 · Cited by 1
Evaluation of Multi-Model Architecture Against Single-Model PPO in the MiniGrid Environment
Ciganović, I., Pluskoski, A., Jovanović, M., & Vasiljevic, J. (2026) · Electronics · doi:10.3390/electronics15143170 · Cited by 2
“It improves learning efficiency by guiding the agent through the learning process, first decomposing the solution into sub-tasks and then ordering them from easier to harder, allowing the agent to gradually acquire increasingly complex behaviors [18].”
Physics‐Embedded End‐To‐End Differentiable Framework for High‐Precision and Robust Interferometry
Shi, R., Yin, P., Wu, B., Zhang, T., & Bai, J. (2026) · Laser & Photonics Review · doi:10.1002/lpor.71606
Lee, W. S., Kwok, I. H., Leong, K. I., et al. (2025) · Asia Conference on Cognitive Engineering and Intelligent Interaction (CEII) · doi:10.1109/ceii67787.2025.00035
“The constraint transitions at epochs 50 and 150 fundamentally alter the learning dynamics, consistent with curriculum learning principles established by Bengio et al. [47] in deep learning contexts.”
Autonomous Obstacle Avoidance and Landing of Quadrotor Using Proximal Policy Optimization
Zhang, J., Li, F., Hong, C., & Lyu, S. (2025) · Asia Conference on Cognitive Engineering and Intelligent Interaction (CEII) · doi:10.1109/ceii67787.2025.00011
“PPO struggles to explore optimal strategies in environments with obstacles, thus we integrated curriculum learning (CL) [17] into the PPO training process.”
Xu, X. (2026) · International Conference on Machine Learning and Intelligent Systems Engineering (MLISE) · doi:10.1109/mlise70044.2026.11607521
Yu, W., Yu, Z., Zuo, J., Yang, W., Ju, Y., & Bao, W. (2026) · ACM Transactions on Asian and Low-Resource Language Information Processing · doi:10.1145/3830466
Visual Analytics for Understanding Curriculum Learning and Generalization in PPO Agents
Zhang, Q., & Reppen, A. (2026) · Conference proceedings - Canadian Conference on Electrical and Computer Engineering · doi:10.1109/ccece68150.2026.11609982
“In practice, training RL agents for non-trivial tasks often relies on curriculum learning, in which agents are exposed to progressively increasing levels of task difficulty to facilitate learning and convergence [3], [4].”
Applications and challenges of machine learning in metal additive manufacturing
Peng, X., Xie, Z., Liu, Y., Hu, P., Zhang, L., Wang, M., & Ke, W. (2026) · Materials Research Letters · doi:10.1080/21663831.2026.2702018
Twitch: Learning Abstractions for Equational Theorem Proving
Axelrod, G., Johansson, M., & Smallbone, N. (2026) · Lecture notes in computer science · doi:10.1007/978-3-032-32589-1_4
Jing, H., Zhang, J., Yang, F., et al. (2026) · IEEE Systems Journal · doi:10.1109/jsyst.2026.3706992
“The sequence of attacks follows the principle of progressing from easy to difficult, grounded in curriculum learning [28].”
Qu, Y., Feng, X., Luo, L., Deng, X., Kang, R., Lv, G., Shi, J., Peng, W., Gan, J., Cai, K., et al. (2026) · Biomimetics · doi:10.3390/biomimetics11080526 · Cited by 1
“This staged approach helps gradually integrate physical constraints while ensuring that the model maintains a solid understanding of the underlying data structure [33].”
Guo, H., Zhang, Q., Liang, X., et al. (2026) · International Conference on Computer and Communication Systems (ICCCS) · doi:10.1109/icccs69761.2026.11613487
Zero-Shot, Standard Fine-Tuning, and Curriculum Learning Approaches for VQA in GI Endoscopy
Azmoodeh-Kalati, M., Maghareh, M. S., & Lashgari, R. (2026) · International Interdisciplinary Conference on Artificial Intelligence: Engineering, Health, Finance and Humanities (IICAI) · doi:10.1109/iicai70155.2026.11620860
“Bengio et al. (2009) formalized the curriculum learning paradigm, demonstrating faster convergence and improved generalization when progressively ordered data is used [10].”
BARAKAT, M., Zhu, X., & Xu, H. (2026) · SSRN Electronic Journal · doi:10.2139/ssrn.7198427
Ye, K., Wu, Z., Wu, Y., Qian, H., Woo, S., Srigrarom, S. S., & Khoo, B. C. (2026) · Aerospace · doi:10.3390/aerospace13080680 · Cited by 1
Knowledge-guided polygon vertex learning with Vision Transformer for tumor target delineation
Fan, Y., Zhou, X., Chen, A., et al. (2026) · Knowledge-Based Systems · doi:10.1016/j.knosys.2026.116711
Gong, S., Wang, M., Wang, J., & Cai, Y. (2026) · IEEE International Conference on Program Comprehension (ICPC) · doi:10.1145/3794763.3794823 · Cited by 1
“CLCoSum [17] further leveraged data augmentation and curriculum learning [3] to enhance code summarization performance.”
Atlas Synchronization in the Hierarchical Federated Learning Continuum
Iosifidis, A., Karagiannis, V., & Schulte, S. (2026) · IEEE International Conference on Fog and Edge Computing (ICFEC) · doi:10.1109/icfec69006.2026.00018
“Some designs incorporate curriculum learning and introduce void classes progressively based on semantic proximity or estimated transferability [6].”
InfluentialCLRe: A Synergistic Dual‐Engine Framework for One‐Step Retrosynthesis Prediction
Su, T.-H., Wang, X.-T., Li, M.-S., Qin, G.-H., Hu, S.-B., & Zhang, T.-Y. (2026) · Advancement of science · doi:10.1002/advs.76827
“While Curriculum Learning (CL) [ 31–34 ] offers a theoretical solution by organizing materials from “easy to hard,” its application in chemistry is hindered by the lack of a label-free, computationally efficient metric to quantify the context-dependent difficulty of molecular transformations.”
Learning Compressed AIS Trajectories with VQ-VAE for Activity Classification
Chronis, C., Chatziargyros, S., Eirinaki, M., Tserpes, K., & Varlamis, I. (2026) · IEEE International Conference on Mobile Data Management (MDM) · doi:10.1109/mdm71479.2026.00080
“The warm-up weight β ( t ) , inspired by curriculum-style training [1], delays strong classification supervision to avoid early majority-class bias.”
Dom4LLM: Domain-Adaptive Large Language Models for Automated Code Clone Refactoring
Xia, M., & Song, M. (2026) · IEEE/ACIS International Conference on Software Engineering Research, Management and Applications (SERA) · doi:10.1109/sera69989.2026.11618621
“Curriculum learning (CL), introduced by [9], improves convergence by training on increasingly difficult data.”
Ma, T., Li, S., Tian, J., et al. (2026) · IEEE Geoscience and Remote Sensing Letters · doi:10.1109/lgrs.2026.3718559
“Incremental curriculum learning [20] is used only during training to avoid unstable optimization caused by di ffi cult samples.”
Siddika, F., Hossen, M. A., Zhang, W., et al. (2026) · IEEE International Symposium on Cluster, Cloud and Internet Computing (CCGrid) · doi:10.1109/ccgrid68966.2026.00050
“The final importance score I i is computed as: where k controls the sigmoid steepness to ensure smoothness, following Curriculum Learning [18] principles.”
Semi-supervised Domain Adaptation with Entorpy-Guided Curriculum and Contrastive Learning
Hwang, S., & Kim, S. B. (2026) · Lecture notes in computer science · doi:10.1007/978-3-032-31663-9_4
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.