ICLR, NeurIPS & ICML Conference Analysis: Authors and Citations by Acceptance Type
This analysis examines three of the most prestigious machine learning conferences—ICLR (International Conference on Learning Representations), NeurIPS (Neural Information Processing Systems), and ICML (International Conference on Machine Learning)—to understand how author counts and citation patterns vary across different acceptance types (Oral, Spotlight, Poster, and Rejected papers).
1. Author Count Analysis
We first examine how the number of authors varies across different acceptance types. This can reveal interesting patterns about collaboration in accepted vs. rejected papers.
Key Finding: The analysis shows that accepted papers (Oral, Spotlight, Poster) tend to have similar author counts, while rejected papers may show different patterns. Collaboration appears to be important across all acceptance types.
1.1 Median Author Count Analysis
While mean author counts provide one perspective, median values can be more robust to outliers and better represent typical collaboration patterns. Below we examine median author counts across conferences and acceptance types.
Key Finding: Comparing mean and median author counts reveals whether collaboration patterns are consistent or if there are outliers (e.g., very large author teams) that skew the averages. The median provides a more typical view of collaboration size.
ICLR Statistics
| Status | Count | Avg Authors | Median Authors | Avg Citations | Median Citations | Max Citations |
|---|---|---|---|---|---|---|
| Oral | 490 | 5.46 | 5.0 | 452.9 | 23.5 | 62,867 |
| Spotlight | 1,143 | 5.47 | 5.0 | 84.0 | 15.0 | 8,832 |
| Poster | 9,290 | 5.01 | 4.0 | 121.6 | 12.0 | 47,498 |
| Reject | 17,379 | 4.46 | 4.0 | 23.7 | 1.0 | 19,629 |
Additional Insights for ICLR:
• 244 papers (1.1%) have 1,000+ citations
• 511 papers (2.3%) have 500+ citations
• 2,262 papers (10.2%) have 100+ citations
• Average authors per paper: 5.01
• Papers span years 2013 to 2026
• 244 papers (1.1%) have 1,000+ citations
• 511 papers (2.3%) have 500+ citations
• 2,262 papers (10.2%) have 100+ citations
• Average authors per paper: 5.01
• Papers span years 2013 to 2026
NeurIPS Statistics
| Status | Count | Avg Authors | Median Authors | Avg Citations | Median Citations | Max Citations |
|---|---|---|---|---|---|---|
| Oral | 610 | 5.00 | 4.0 | 390.4 | 54.0 | 47,439 |
| Spotlight | 2,685 | 4.83 | 4.0 | 192.9 | 13.0 | 184,033 |
| Poster | 27,550 | 4.47 | 4.0 | 109.8 | 14.0 | 82,368 |
| Reject | 1,213 | 5.07 | 4.0 | 10.9 | 0.0 | 1,293 |
Additional Insights for NeurIPS:
• 532 papers (1.7%) have 1,000+ citations
• 1,095 papers (3.5%) have 500+ citations
• 5,093 papers (16.5%) have 100+ citations
• Average authors per paper: 4.50
• Papers span years 1987 to 2025
• 532 papers (1.7%) have 1,000+ citations
• 1,095 papers (3.5%) have 500+ citations
• 5,093 papers (16.5%) have 100+ citations
• Average authors per paper: 4.50
• Papers span years 1987 to 2025
ICML Statistics
| Status | Count | Avg Authors | Median Authors | Avg Citations | Median Citations | Max Citations |
|---|---|---|---|---|---|---|
| Oral | 2,098 | 4.18 | 4.0 | 213.1 | 48.0 | 35,305 |
| Spotlight | 2,548 | 4.40 | 4.0 | 66.8 | 24.0 | 8,716 |
| Poster | 9,684 | 4.53 | 4.0 | 90.1 | 6.0 | 62,227 |
| Reject | 185 | 4.59 | 4.0 | -0.3 | 0.0 | 0 |
Additional Insights for ICML:
• 217 papers (1.5%) have 1,000+ citations
• 451 papers (3.1%) have 500+ citations
• 2,097 papers (14.6%) have 100+ citations
• Average authors per paper: 4.45
• Papers span years 2013 to 2025
• 217 papers (1.5%) have 1,000+ citations
• 451 papers (3.1%) have 500+ citations
• 2,097 papers (14.6%) have 100+ citations
• Average authors per paper: 4.45
• Papers span years 2013 to 2025
2. Citation Analysis
Citation counts are a key metric of research impact. We analyze how citations vary by acceptance type, which can provide insights into the relationship between peer review decisions and long-term impact.
Key Finding: Oral presentations typically receive the highest citations, followed by Spotlight and Poster presentations. However, there are notable exceptions, and citation patterns evolve over time as papers accumulate citations.
3. Top Cited Papers by Year
Below we highlight the most highly cited papers from each year in both conferences. These papers represent some of the most influential work in machine learning and have had significant impact on the field.
ICLR - Top Cited Papers by Year
Year 2026
-
#1: CyberV: A Cybernetic Framework for Enhancing Logical Reasoning in Video Understanding
-
#2: SMART-3D: Scaling Masked AutoRegressive Transformer for Efficient 3D Shape Generation
-
#3: SSTP: Efficient Sample Selection for Trajectory Prediction
Year 2025
-
#1: KAN: Kolmogorov–Arnold Networks
[OpenReview] -
#2: SAM 2: Segment Anything in Images and Videos
[OpenReview] -
#3: WizardMath: Empowering Mathematical Reasoning for Large Language Models via Reinforced Evol-Instruct
[OpenReview]
Year 2024
-
#1: YOLOV6: A SINGLE-STAGE OBJECT DETECTION FRAMEWORK FOR INDUSTRIAL APPLICATIONS
-
#2: MiniGPT-4: Enhancing Vision-Language Understanding with Advanced Large Language Models
[OpenReview] -
#3: SDXL: Improving Latent Diffusion Models for High-Resolution Image Synthesis
[OpenReview]
Year 2023
-
#1: ReAct: Synergizing Reasoning and Acting in Language Models
[OpenReview] -
#2: DreamFusion: Text-to-3D using 2D Diffusion
[OpenReview] -
#3: DINO: DETR with Improved DeNoising Anchor Boxes for End-to-End Object Detection
[OpenReview]
Year 2022
-
#1: LoRA: Low-Rank Adaptation of Large Language Models
[OpenReview] -
#2: Finetuned Language Models are Zero-Shot Learners
[OpenReview] -
#3: BEiT: BERT Pre-Training of Image Transformers
[OpenReview]
Year 2021
-
#1: An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale
[OpenReview] -
#2: Denoising Diffusion Implicit Models
[OpenReview] -
#3: Score-Based Generative Modeling through Stochastic Differential Equations
[OpenReview]
Year 2020
-
#1: ALBERT: A Lite BERT for Self-supervised Learning of Language Representations
-
#2: BERTScore: Evaluating Text Generation with BERT
-
#3: ELECTRA: Pre-training Text Encoders as Discriminators Rather Than Generators
Year 2019
-
#1: Decoupled Weight Decay Regularization
[OpenReview] -
#2: How Powerful are Graph Neural Networks?
[OpenReview] -
#3: GLUE: A Multi-Task Benchmark and Analysis Platform for Natural Language Understanding
[OpenReview]
Year 2018
-
#1: Towards Deep Learning Models Resistant to Adversarial Attacks
[OpenReview] -
#2: Graph Attention Networks
[OpenReview] -
#3: mixup: Beyond Empirical Risk Minimization
[OpenReview]
Year 2017
-
#1: Semi-Supervised Classification with Graph Convolutional Networks
-
#2: SGDR: Stochastic Gradient Descent with Warm Restarts
-
#3: Adversarial examples in the physical world
Year 2014
-
#1: Auto-Encoding Variational Bayes
-
#2: Intriguing properties of neural networks
-
#3: Network In Network
Year 2013
-
#1: Efficient Estimation of Word Representations in Vector Space
-
#2: Deep Learning for Detecting Robotic Grasps
-
#3: Zero-Shot Learning Through Cross-Modal Transfer
NeurIPS - Top Cited Papers by Year
Year 2025
-
#1: How Well Can Differential Privacy Be Audited in One Run?
-
#2: Uncertainty-Sensitive Privileged Learning
-
#3: KL Penalty Control via Perturbation for Direct Preference Optimization
Year 2024
-
#1: YOLOv10: Real-Time End-to-End Object Detection
[OpenReview] -
#2: VMamba: Visual State Space Model
[OpenReview] -
#3: CogVLM: Visual Expert for Pretrained Language Models
[OpenReview]
Year 2023
-
#1: Visual Instruction Tuning
[OpenReview] -
#2: Judging LLM-as-a-Judge with MT-Bench and Chatbot Arena
[OpenReview] -
#3: Direct Preference Optimization: Your Language Model is Secretly a Reward Model
[OpenReview]
Year 2022
-
#1: Training language models to follow instructions with human feedback
[OpenReview] -
#2: Chain-of-Thought Prompting Elicits Reasoning in Large Language Models
[OpenReview] -
#3: Photorealistic Text-to-Image Diffusion Models with Deep Language Understanding
[OpenReview]
Year 2021
-
#1: Diffusion Models Beat GANs on Image Synthesis
[OpenReview] -
#2: SegFormer: Simple and Efficient Design for Semantic Segmentation with Transformers
[OpenReview] -
#3: MLP-Mixer: An all-MLP Architecture for Vision
[OpenReview]
Year 2020
-
#1: Language Models are Few-Shot Learners
-
#2: Denoising Diffusion Probabilistic Models
-
#3: Retrieval-Augmented Generation for Knowledge-Intensive NLP Tasks
Year 2019
-
#1: PyTorch: An Imperative Style, High-Performance Deep Learning Library
-
#2: XLNet: Generalized Autoregressive Pretraining for Language Understanding
-
#3: Generative Modeling by Estimating Gradients of the Data Distribution
Year 2018
-
#1: Neural Ordinary Differential Equations
-
#2: CatBoost: unbiased boosting with categorical features
-
#3: Neural Tangent Kernel: Convergence and Generalization in Neural Networks
Year 2017
-
#1: Attention is All you Need
-
#2: A Unified Approach to Interpreting Model Predictions
-
#3: Inductive Representation Learning on Large Graphs
Year 2016
-
#1: Improved Techniques for Training GANs
-
#2: Convolutional Neural Networks on Graphs with Fast Localized Spectral Filtering
-
#3: Matching Networks for One Shot Learning
Year 2015
-
#1: Faster R-CNN: Towards Real-Time Object Detection with Region Proposal Networks
-
#2: Spatial Transformer Networks
-
#3: Convolutional LSTM Network: A Machine Learning Approach for Precipitation Nowcasting
Year 2014
-
#1: Generative Adversarial Nets
-
#2: Sequence to Sequence Learning with Neural Networks
-
#3: How transferable are features in deep neural networks?
Year 2013
-
#1: Distributed Representations of Words and Phrases and their Compositionality
-
#2: Translating Embeddings for Modeling Multi-relational Data
-
#3: Sinkhorn Distances: Lightspeed Computation of Optimal Transport
Year 2012
-
#1: ImageNet Classification with Deep Convolutional Neural Networks
-
#2: Practical Bayesian Optimization of Machine Learning Algorithms
-
#3: Large Scale Distributed Deep Networks
Year 2011
-
#1: Algorithms for Hyper-Parameter Optimization
-
#2: Efficient Inference in Fully Connected CRFs with Gaussian Edge Potentials
-
#3: Hogwild!: A Lock-Free Approach to Parallelizing Stochastic Gradient Descent
Year 2010
-
#1: Efficient and Robust Feature Selection via Joint ℓ2,1-Norms Minimization
-
#2: Double Q-learning
-
#3: Online Learning for Latent Dirichlet Allocation
Year 2009
-
#1: Reading Tea Leaves: How Humans Interpret Topic Models
-
#2: Robust Principal Component Analysis: Exact Recovery of Corrupted Low-Rank Matrices via Convex Opt...
-
#3: Fast Image Deconvolution using Hyper-Laplacian Priors
Year 2008
-
#1: Spectral Hashing
-
#2: Mixed Membership Stochastic Blockmodels
-
#3: Near-optimal Regret Bounds for Reinforcement Learning
Year 2007
-
#1: Probabilistic Matrix Factorization
-
#2: Random Features for Large-Scale Kernel Machines
-
#3: Supervised Topic Models
Year 2006
-
#1: Greedy Layer-Wise Training of Deep Networks
-
#2: Graph-Based Visual Saliency
-
#3: Efficient sparse coding algorithms
Year 2005
-
#1: Distance Metric Learning for Large Margin Nearest Neighbor Classification
-
#2: Laplacian Score for Feature Selection
-
#3: Sparse Gaussian Processes using Pseudo-inputs
Year 2004
-
#1: Sharing Clusters among Related Groups: Hierarchical Dirichlet Processes
-
#2: Neighbourhood Components Analysis
-
#3: Self-Tuning Spectral Clustering
Year 2003
-
#1: Locality Preserving Projections
-
#2: Learning with Local and Global Consistency
-
#3: Online Passive-Aggressive Algorithms
Year 2002
-
#1: Distance Metric Learning with Application to Clustering with Side-Information
-
#2: Stochastic Neighbor Embedding
-
#3: Support Vector Machines for Multiple-Instance Learning
Year 2001
-
#1: Latent Dirichlet Allocation
-
#2: On Spectral Clustering: Analysis and an algorithm
-
#3: Laplacian Eigenmaps and Spectral Techniques for Embedding and Clustering
Year 2000
-
#1: A Neural Probabilistic Language Model
-
#2: Algorithms for Non-negative Matrix Factorization
-
#3: Using the Nyström Method to Speed Up Kernel Machines
Year 1999
-
#1: Policy Gradient Methods for Reinforcement Learning with Function Approximation
-
#2: Actor-Critic Algorithms
-
#3: Support Vector Method for Novelty Detection
Year 1998
-
#1: Exploiting Generative Models in Discriminative Classifiers
-
#2: Kernel PCA and De-Noising in Feature Spaces
-
#3: Semi-Supervised Support Vector Machines
Year 1997
-
#1: Classification by Pairwise Coupling
-
#2: A Framework for Multiple-Instance Learning
-
#3: EM Algorithms for PCA and SPCA
Year 1996
-
#1: Support Vector Regression Machines
-
#2: Support Vector Method for Function Approximation, Regression Estimation and Signal Processing
-
#3: Analysis of Temporal-Diffference Learning with Function Approximation
Year 1995
-
#1: A New Learning Algorithm for Blind Signal Separation
-
#2: Independent Component Analysis of Electroencephalographic Data
-
#3: Gaussian Processes for Regression
Year 1994
-
#1: Neural Network Ensembles, Cross Validation, and Active Learning
-
#2: A Growing Neural Gas Network Learns Topologies
-
#3: Active Learning with Statistical Models
Year 1993
-
#1: Signature Verification using a "Siamese" Time Delay Neural Network
-
#2: Autoencoders, Minimum Description Length and Helmholtz Free Energy
-
#3: Convergence of Stochastic Iterative Dynamic Programming Algorithms
Year 1992
-
#1: Second order derivatives for network pruning: Optimal Brain Surgeon
-
#2: Feudal Reinforcement Learning
-
#3: An Information-Theoretic Approach to Deciphering the Hippocampal Code
Year 1991
-
#1: A Simple Weight Decay Can Improve Generalization
-
#2: Principles of Risk Minimization for Learning Theory
-
#3: Practical Issues in Temporal Difference Learning
Year 1990
-
#1: Generalization by Weight-Elimination with Application to Forecasting
-
#2: SEXNET: A Neural Network Identifies Sex From Human Faces
-
#3: Back Propagation is Sensitive to Initial Conditions
Year 1989
-
#1: Handwritten Digit Recognition with a Back-Propagation Network
-
#2: Optimal Brain Damage
-
#3: The Cascade-Correlation Learning Architecture
Year 1988
-
#1: ALVINN: An Autonomous Land Vehicle in a Neural Network
-
#2: What Size Net Gives Valid Generalization?
-
#3: Training a 3-Node Neural Network is NP-Complete
Year 1987
-
#1: Generalization of Back propagation to Recurrent and Higher Order Neural Networks
-
#2: How Neural Nets Work
-
#3: Supervised Learning of Probability Distributions by Neural Networks
ICML - Top Cited Papers by Year
Year 2025
-
#1: Learning to (Learn at Test Time): RNNs with Expressive Hidden States
[OpenReview] -
#2: AdvPrompter: Fast Adaptive Adversarial Prompting for LLMs
[OpenReview] -
#3: rStar-Math: Small LLMs Can Master Math Reasoning with Self-Evolved Deep Thinking
[OpenReview]
Year 2024
-
#1: Scaling Rectified Flow Transformers for High-Resolution Image Synthesis
[OpenReview] -
#2: MM-Vet: Evaluating Large Multimodal Models for Integrated Capabilities
[OpenReview] -
#3: Improving Factuality and Reasoning in Language Models through Multiagent Debate
[OpenReview]
Year 2023
-
#1: BLIP-2: Bootstrapping Language-Image Pre-training with Frozen Image Encoders and Large Language M...
[OpenReview] -
#2: Robust Speech Recognition via Large-Scale Weak Supervision
[OpenReview] -
#3: PaLM-E: An Embodied Multimodal Language Model
[OpenReview]
Year 2022
-
#1: BLIP: Bootstrapping Language-Image Pre-training for Unified Vision-Language Understanding and Gen...
-
#2: GLIDE: Towards Photorealistic Image Generation and Editing with Text-Guided Diffusion Models
-
#3: FEDformer: Frequency Enhanced Decomposed Transformer for Long-term Series Forecasting
Year 2021
-
#1: Learning Transferable Visual Models From Natural Language Supervision
-
#2: Training data-efficient image transformers & distillation through attention
-
#3: Zero-Shot Text-to-Image Generation
Year 2020
-
#1: A Simple Framework for Contrastive Learning of Visual Representations
-
#2: SCAFFOLD: Stochastic Controlled Averaging for Federated Learning
-
#3: PEGASUS: Pre-training with Extracted Gap-sentences for Abstractive Summarization
Year 2019
-
#1: EfficientNet: Rethinking Model Scaling for Convolutional Neural Networks
-
#2: Self-Attention Generative Adversarial Networks
-
#3: Parameter-Efficient Transfer Learning for NLP
Year 2018
-
#1: Soft Actor-Critic: Off-Policy Maximum Entropy Deep Reinforcement Learning with a Stochastic Actor
-
#2: Addressing Function Approximation Error in Actor-Critic Methods
-
#3: Obfuscated Gradients Give a False Sense of Security: Circumventing Defenses to Adversarial Examples
Year 2017
-
#1: Wasserstein Generative Adversarial Networks
-
#2: Model-Agnostic Meta-Learning for Fast Adaptation of Deep Networks
-
#3: Neural Message Passing for Quantum Chemistry
Year 2016
-
#1: Asynchronous Methods for Deep Reinforcement Learning
-
#2: Dropout as a Bayesian Approximation: Representing Model Uncertainty in Deep Learning
-
#3: Dueling Network Architectures for Deep Reinforcement Learning
Year 2015
-
#1: Batch Normalization: Accelerating Deep Network Training by Reducing Internal Covariate Shift
-
#2: Show, Attend and Tell: Neural Image Caption Generation with Visual Attention
-
#3: Trust Region Policy Optimization
Year 2014
-
#1: Distributed Representations of Sentences and Documents
-
#2: Stochastic Backpropagation and Approximate Inference in Deep Generative Models
-
#3: DeCAF: A Deep Convolutional Activation Feature for Generic Visual Recognition
Year 2013
-
#1: On the difficulty of training recurrent neural networks
-
#2: On the importance of initialization and momentum in deep learning
-
#3: Regularization of Neural Networks using DropConnect
4. Combined Analysis
By comparing ICLR, NeurIPS, and ICML together, we can see broader trends in the machine learning research community. All three conferences show similar patterns, suggesting these findings are generalizable across top-tier ML venues.
Conclusion
This analysis reveals important patterns in how acceptance type relates to both collaboration (author counts) and impact (citations). While oral presentations tend to receive more citations on average, poster presentations make up the majority of accepted papers and include many highly-cited works. The data suggests that acceptance type is not the sole determinant of future impact, and that collaboration (as measured by author count) is important across all acceptance types.
Note: Citation counts are from Google Scholar and represent cumulative citations up to the time of data collection. Older papers have had more time to accumulate citations, which may affect the analysis.