[ { "canonical": true, "canonical_title": "A Logical Calculus of the Ideas Immanent in Nervous Activity", "conflict_flag": false, "editorial": { "authors": "Warren McCulloch, Walter Pitts", "category": "Foundations", "confidence": "High", "significance": "First mathematical model of the neuron; the conceptual origin of neural networks.", "venue": "Bulletin of Mathematical Biophysics" }, "id": "paper-0001", "language": "en", "level": null, "original_title": null, "work_type": "paper", "year": 1943 }, { "canonical": true, "canonical_title": "As We May Think", "conflict_flag": false, "editorial": { "authors": "Vannevar Bush", "category": "Foundations", "confidence": "High", "significance": "The memex vision; founding document of augmenting human intellect with machines.", "venue": "The Atlantic" }, "id": "paper-0002", "language": "en", "level": null, "original_title": null, "work_type": "paper", "year": 1945 }, { "canonical": true, "canonical_title": "A Mathematical Theory of Communication", "conflict_flag": false, "editorial": { "authors": "Claude E. Shannon", "category": "Foundations", "confidence": "High", "significance": "Created information theory; the quantitative substrate of all machine learning.", "venue": "Bell System Technical Journal" }, "id": "paper-0003", "language": "en", "level": null, "original_title": null, "work_type": "paper", "year": 1948 }, { "canonical": true, "canonical_title": "Computing Machinery and Intelligence", "conflict_flag": false, "editorial": { "authors": "Alan M. Turing", "category": "Foundations", "confidence": "High", "significance": "Posed 'can machines think', proposed the imitation game; the field's founding question.", "venue": "Mind" }, "id": "paper-0004", "language": "en", "level": null, "original_title": null, "work_type": "paper", "year": 1950 }, { "canonical": true, "canonical_title": "A Proposal for the Dartmouth Summer Research Project on Artificial Intelligence", "conflict_flag": false, "editorial": { "authors": "John McCarthy, Marvin Minsky, Nathaniel Rochester, Claude Shannon", "category": "Foundations", "confidence": "High", "significance": "Coined 'artificial intelligence' and framed the research programme.", "venue": "Proposal (Dartmouth)" }, "id": "paper-0005", "language": "en", "level": null, "original_title": null, "work_type": "paper", "year": 1955 }, { "canonical": true, "canonical_title": "The Perceptron: A Probabilistic Model for Information Storage and Organization in the Brain", "conflict_flag": false, "editorial": { "authors": "Frank Rosenblatt", "category": "Foundations", "confidence": "High", "significance": "First trainable neural classifier; launched learning machines and their first hype cycle.", "venue": "Psychological Review" }, "id": "paper-0006", "language": "en", "level": null, "original_title": null, "work_type": "paper", "year": 1958 }, { "canonical": true, "canonical_title": "Some Studies in Machine Learning Using the Game of Checkers", "conflict_flag": false, "editorial": { "authors": "Arthur L. Samuel", "category": "Foundations", "confidence": "High", "significance": "Coined 'machine learning'; first self-improving game program.", "venue": "IBM Journal" }, "id": "paper-0007", "language": "en", "level": null, "original_title": null, "work_type": "paper", "year": 1959 }, { "canonical": true, "canonical_title": "Programs with Common Sense", "conflict_flag": false, "editorial": { "authors": "John McCarthy", "category": "Foundations", "confidence": "High", "significance": "The Advice Taker; founding statement of logic-based AI and knowledge representation.", "venue": "Mechanisation of Thought Processes" }, "id": "paper-0008", "language": "en", "level": null, "original_title": null, "work_type": "paper", "year": 1959 }, { "canonical": true, "canonical_title": "Man-Computer Symbiosis", "conflict_flag": false, "editorial": { "authors": "J. C. R. Licklider", "category": "Foundations", "confidence": "High", "significance": "The augmentation-versus-automation agenda that still structures AI debates.", "venue": "IRE Transactions on Human Factors" }, "id": "paper-0009", "language": "en", "level": null, "original_title": null, "work_type": "paper", "year": 1960 }, { "canonical": true, "canonical_title": "Steps Toward Artificial Intelligence", "conflict_flag": false, "editorial": { "authors": "Marvin Minsky", "category": "Foundations", "confidence": "High", "significance": "Early synthesis of search, learning, and planning as the components of AI.", "venue": "Proceedings of the IRE" }, "id": "paper-0010", "language": "en", "level": null, "original_title": null, "work_type": "paper", "year": 1961 }, { "canonical": true, "canonical_title": "GPS: A Program That Simulates Human Thought", "conflict_flag": false, "editorial": { "authors": "Allen Newell, Herbert A. Simon", "category": "Foundations", "confidence": "High", "significance": "General Problem Solver; means-ends analysis and the symbolic cognition paradigm.", "venue": "Lernende Automaten / RAND" }, "id": "paper-0011", "language": "en", "level": null, "original_title": null, "work_type": "paper", "year": 1961 }, { "canonical": true, "canonical_title": "Fuzzy Sets", "conflict_flag": false, "editorial": { "authors": "Lotfi A. Zadeh", "category": "Foundations", "confidence": "Medium", "significance": "Founded fuzzy logic; a major non-probabilistic approach to reasoning under vagueness.", "venue": "Information and Control" }, "id": "paper-0012", "language": "en", "level": null, "original_title": null, "work_type": "paper", "year": 1965 }, { "canonical": true, "canonical_title": "ELIZA: A Computer Program for the Study of Natural Language Communication", "conflict_flag": false, "editorial": { "authors": "Joseph Weizenbaum", "category": "Foundations", "confidence": "High", "significance": "First chatbot; revealed the human tendency to project understanding onto machines.", "venue": "Communications of the ACM" }, "id": "paper-0013", "language": "en", "level": null, "original_title": null, "work_type": "paper", "year": 1966 }, { "canonical": true, "canonical_title": "Some Philosophical Problems from the Standpoint of Artificial Intelligence", "conflict_flag": false, "editorial": { "authors": "John McCarthy, Patrick J. Hayes", "category": "Foundations", "confidence": "High", "significance": "Introduced the frame problem and situation calculus.", "venue": "Machine Intelligence 4" }, "id": "paper-0014", "language": "en", "level": null, "original_title": null, "work_type": "paper", "year": 1969 }, { "canonical": true, "canonical_title": "STRIPS: A New Approach to the Application of Theorem Proving to Problem Solving", "conflict_flag": false, "editorial": { "authors": "Richard E. Fikes, Nils J. Nilsson", "category": "Foundations", "confidence": "High", "significance": "The planning formalism that dominated automated planning for decades.", "venue": "Artificial Intelligence" }, "id": "paper-0015", "language": "en", "level": null, "original_title": null, "work_type": "paper", "year": 1971 }, { "canonical": true, "canonical_title": "Adaptation in Natural and Artificial Systems (foundational monograph)", "conflict_flag": false, "editorial": { "authors": "John H. Holland", "category": "Foundations", "confidence": "High", "significance": "Genetic algorithms; evolution as a computational search paradigm.", "venue": "University of Michigan Press" }, "id": "paper-0016", "language": "en", "level": null, "original_title": null, "work_type": "paper", "year": 1975 }, { "canonical": true, "canonical_title": "Maximum Likelihood from Incomplete Data via the EM Algorithm", "conflict_flag": false, "editorial": { "authors": "Arthur Dempster, Nan Laird, Donald Rubin", "category": "Statistical Learning", "confidence": "High", "significance": "The EM algorithm; workhorse of latent-variable estimation.", "venue": "Journal of the Royal Statistical Society B" }, "id": "paper-0017", "language": "en", "level": null, "original_title": null, "work_type": "paper", "year": 1977 }, { "canonical": true, "canonical_title": "Minds, Brains, and Programs", "conflict_flag": false, "editorial": { "authors": "John R. Searle", "category": "Foundations", "confidence": "High", "significance": "The Chinese Room argument; the canonical philosophical attack on strong AI.", "venue": "Behavioral and Brain Sciences" }, "id": "paper-0018", "language": "en", "level": null, "original_title": null, "work_type": "paper", "year": 1980 }, { "canonical": true, "canonical_title": "Neocognitron: A Self-Organizing Neural Network Model", "conflict_flag": false, "editorial": { "authors": "Kunihiko Fukushima", "category": "Neural Networks (Classic)", "confidence": "High", "significance": "The architectural ancestor of convolutional networks.", "venue": "Biological Cybernetics" }, "id": "paper-0019", "language": "en", "level": null, "original_title": null, "work_type": "paper", "year": 1980 }, { "canonical": true, "canonical_title": "Neural Networks and Physical Systems with Emergent Collective Computational Abilities", "conflict_flag": false, "editorial": { "authors": "John J. Hopfield", "category": "Neural Networks (Classic)", "confidence": "High", "significance": "Hopfield networks; energy-based associative memory that revived the field (2024 Nobel).", "venue": "PNAS" }, "id": "paper-0020", "language": "en", "level": null, "original_title": null, "work_type": "paper", "year": 1982 }, { "canonical": true, "canonical_title": "Self-Organized Formation of Topologically Correct Feature Maps", "conflict_flag": false, "editorial": { "authors": "Teuvo Kohonen", "category": "Neural Networks (Classic)", "confidence": "Medium", "significance": "Self-organizing maps; unsupervised topology-preserving learning.", "venue": "Biological Cybernetics" }, "id": "paper-0021", "language": "en", "level": null, "original_title": null, "work_type": "paper", "year": 1982 }, { "canonical": true, "canonical_title": "A Learning Algorithm for Boltzmann Machines", "conflict_flag": false, "editorial": { "authors": "David H. Ackley, Geoffrey E. Hinton, Terrence J. Sejnowski", "category": "Neural Networks (Classic)", "confidence": "High", "significance": "Stochastic networks and unsupervised learning of internal representations.", "venue": "Cognitive Science" }, "id": "paper-0022", "language": "en", "level": null, "original_title": null, "work_type": "paper", "year": 1985 }, { "canonical": true, "canonical_title": "A Robust Layered Control System for a Mobile Robot", "conflict_flag": false, "editorial": { "authors": "Rodney A. Brooks", "category": "Foundations", "confidence": "High", "significance": "Subsumption architecture; behaviour-based robotics against the symbolic mainstream.", "venue": "IEEE Journal of Robotics and Automation" }, "id": "paper-0023", "language": "en", "level": null, "original_title": null, "work_type": "paper", "year": 1986 }, { "canonical": true, "canonical_title": "Learning Representations by Back-Propagating Errors", "conflict_flag": false, "editorial": { "authors": "David E. Rumelhart, Geoffrey E. Hinton, Ronald J. Williams", "category": "Neural Networks (Classic)", "confidence": "High", "significance": "Made backpropagation the standard training algorithm; the engine of all deep learning.", "venue": "Nature" }, "id": "paper-0024", "language": "en", "level": null, "original_title": null, "work_type": "paper", "year": 1986 }, { "canonical": true, "canonical_title": "Fusion, Propagation, and Structuring in Belief Networks", "conflict_flag": false, "editorial": { "authors": "Judea Pearl", "category": "Statistical Learning", "confidence": "High", "significance": "Bayesian networks; principled probabilistic reasoning in AI (Turing Award work).", "venue": "Artificial Intelligence" }, "id": "paper-0025", "language": "en", "level": null, "original_title": null, "work_type": "paper", "year": 1986 }, { "canonical": true, "canonical_title": "Learning to Predict by the Methods of Temporal Differences", "conflict_flag": false, "editorial": { "authors": "Richard S. Sutton", "category": "Reinforcement Learning", "confidence": "High", "significance": "TD learning; the core idea of modern RL.", "venue": "Machine Learning" }, "id": "paper-0026", "language": "en", "level": null, "original_title": null, "work_type": "paper", "year": 1988 }, { "canonical": true, "canonical_title": "Multilayer Feedforward Networks Are Universal Approximators", "conflict_flag": false, "editorial": { "authors": "Kurt Hornik, Maxwell Stinchcombe, Halbert White", "category": "Neural Networks (Classic)", "confidence": "High", "significance": "Proved neural nets can approximate any function; the theoretical licence for the field.", "venue": "Neural Networks" }, "id": "paper-0027", "language": "en", "level": null, "original_title": null, "work_type": "paper", "year": 1989 }, { "canonical": true, "canonical_title": "A Tutorial on Hidden Markov Models and Selected Applications in Speech Recognition", "conflict_flag": false, "editorial": { "authors": "Lawrence R. Rabiner", "category": "Statistical Learning", "confidence": "Medium", "significance": "Made HMMs the standard sequential model for two decades of speech and NLP.", "venue": "Proceedings of the IEEE" }, "id": "paper-0028", "language": "en", "level": null, "original_title": null, "work_type": "paper", "year": 1989 }, { "canonical": true, "canonical_title": "The Symbol Grounding Problem", "conflict_flag": false, "editorial": { "authors": "Stevan Harnad", "category": "Foundations", "confidence": "Medium", "significance": "Named the problem of how symbols acquire meaning; resurfaces in every LLM debate.", "venue": "Physica D" }, "id": "paper-0029", "language": "en", "level": null, "original_title": null, "work_type": "paper", "year": 1990 }, { "canonical": true, "canonical_title": "Finding Structure in Time", "conflict_flag": false, "editorial": { "authors": "Jeffrey L. Elman", "category": "Neural Networks (Classic)", "confidence": "Medium", "significance": "Simple recurrent networks; sequence learning before LSTMs.", "venue": "Cognitive Science" }, "id": "paper-0030", "language": "en", "level": null, "original_title": null, "work_type": "paper", "year": 1990 }, { "canonical": true, "canonical_title": "Intelligence Without Representation", "conflict_flag": false, "editorial": { "authors": "Rodney A. Brooks", "category": "Foundations", "confidence": "High", "significance": "Manifesto for embodied intelligence; the strongest internal critique of GOFAI.", "venue": "Artificial Intelligence" }, "id": "paper-0031", "language": "en", "level": null, "original_title": null, "work_type": "paper", "year": 1991 }, { "canonical": true, "canonical_title": "Q-Learning", "conflict_flag": false, "editorial": { "authors": "Christopher J. C. H. Watkins, Peter Dayan", "category": "Reinforcement Learning", "confidence": "High", "significance": "Model-free value learning with convergence proof.", "venue": "Machine Learning" }, "id": "paper-0032", "language": "en", "level": null, "original_title": null, "work_type": "paper", "year": 1992 }, { "canonical": true, "canonical_title": "Simple Statistical Gradient-Following Algorithms for Connectionist Reinforcement Learning", "conflict_flag": false, "editorial": { "authors": "Ronald J. Williams", "category": "Reinforcement Learning", "confidence": "Medium", "significance": "REINFORCE; the original policy-gradient method.", "venue": "Machine Learning" }, "id": "paper-0033", "language": "en", "level": null, "original_title": null, "work_type": "paper", "year": 1992 }, { "canonical": true, "canonical_title": "Causal Diagrams for Empirical Research", "conflict_flag": false, "editorial": { "authors": "Judea Pearl", "category": "Statistical Learning", "confidence": "High", "significance": "The do-calculus; the formal foundation of modern causal inference.", "venue": "Biometrika" }, "id": "paper-0034", "language": "en", "level": null, "original_title": null, "work_type": "paper", "year": 1995 }, { "canonical": true, "canonical_title": "Support-Vector Networks", "conflict_flag": false, "editorial": { "authors": "Corinna Cortes, Vladimir Vapnik", "category": "Statistical Learning", "confidence": "High", "significance": "SVMs; the dominant classifier of the pre-deep-learning era.", "venue": "Machine Learning" }, "id": "paper-0035", "language": "en", "level": null, "original_title": null, "work_type": "paper", "year": 1995 }, { "canonical": true, "canonical_title": "Temporal Difference Learning and TD-Gammon", "conflict_flag": false, "editorial": { "authors": "Gerald Tesauro", "category": "Reinforcement Learning", "confidence": "High", "significance": "Self-play RL reaches expert backgammon; the proof of concept for everything later.", "venue": "Communications of the ACM" }, "id": "paper-0036", "language": "en", "level": null, "original_title": null, "work_type": "paper", "year": 1995 }, { "canonical": true, "canonical_title": "Regression Shrinkage and Selection via the Lasso", "conflict_flag": false, "editorial": { "authors": "Robert Tibshirani", "category": "Statistical Learning", "confidence": "Medium", "significance": "L1 regularization; sparse models across statistics and ML.", "venue": "Journal of the Royal Statistical Society B" }, "id": "paper-0037", "language": "en", "level": null, "original_title": null, "work_type": "paper", "year": 1996 }, { "canonical": true, "canonical_title": "Long Short-Term Memory", "conflict_flag": false, "editorial": { "authors": "Sepp Hochreiter, Juergen Schmidhuber", "category": "Neural Networks (Classic)", "confidence": "High", "significance": "Solved vanishing gradients for sequences; powered a decade of speech and language AI.", "venue": "Neural Computation" }, "id": "paper-0038", "language": "en", "level": null, "original_title": null, "work_type": "paper", "year": 1997 }, { "canonical": true, "canonical_title": "A Decision-Theoretic Generalization of On-Line Learning and an Application to Boosting", "conflict_flag": false, "editorial": { "authors": "Yoav Freund, Robert E. Schapire", "category": "Statistical Learning", "confidence": "High", "significance": "AdaBoost; proved weak learners can be combined into strong ones.", "venue": "Journal of Computer and System Sciences" }, "id": "paper-0039", "language": "en", "level": null, "original_title": null, "work_type": "paper", "year": 1997 }, { "canonical": true, "canonical_title": "No Free Lunch Theorems for Optimization", "conflict_flag": false, "editorial": { "authors": "David H. Wolpert, William G. Macready", "category": "Statistical Learning", "confidence": "Medium", "significance": "No algorithm wins on all problems; a standing caution against universal claims.", "venue": "IEEE Transactions on Evolutionary Computation" }, "id": "paper-0040", "language": "en", "level": null, "original_title": null, "work_type": "paper", "year": 1997 }, { "canonical": true, "canonical_title": "Gradient-Based Learning Applied to Document Recognition", "conflict_flag": false, "editorial": { "authors": "Yann LeCun, Leon Bottou, Yoshua Bengio, Patrick Haffner", "category": "Neural Networks (Classic)", "confidence": "High", "significance": "LeNet; the canonical demonstration that CNNs work on real tasks.", "venue": "Proceedings of the IEEE" }, "id": "paper-0041", "language": "en", "level": null, "original_title": null, "work_type": "paper", "year": 1998 }, { "canonical": true, "canonical_title": "The PageRank Citation Ranking: Bringing Order to the Web", "conflict_flag": false, "editorial": { "authors": "Lawrence Page, Sergey Brin, Rajeev Motwani, Terry Winograd", "category": "Statistical Learning", "confidence": "Medium", "significance": "Link-based ranking; the algorithm that organized the web's information.", "venue": "Stanford Technical Report" }, "id": "paper-0042", "language": "en", "level": null, "original_title": null, "work_type": "paper", "year": 1998 }, { "canonical": true, "canonical_title": "Random Forests", "conflict_flag": false, "editorial": { "authors": "Leo Breiman", "category": "Statistical Learning", "confidence": "High", "significance": "The most-used classical ML algorithm in applied practice.", "venue": "Machine Learning" }, "id": "paper-0043", "language": "en", "level": null, "original_title": null, "work_type": "paper", "year": 2001 }, { "canonical": true, "canonical_title": "Statistical Modeling: The Two Cultures", "conflict_flag": false, "editorial": { "authors": "Leo Breiman", "category": "Statistical Learning", "confidence": "High", "significance": "Named the split between data modeling and algorithmic prediction; prophetic for ML's rise.", "venue": "Statistical Science" }, "id": "paper-0044", "language": "en", "level": null, "original_title": null, "work_type": "paper", "year": 2001 }, { "canonical": true, "canonical_title": "Greedy Function Approximation: A Gradient Boosting Machine", "conflict_flag": false, "editorial": { "authors": "Jerome H. Friedman", "category": "Statistical Learning", "confidence": "High", "significance": "Gradient boosting; the backbone of tabular ML to this day.", "venue": "Annals of Statistics" }, "id": "paper-0045", "language": "en", "level": null, "original_title": null, "work_type": "paper", "year": 2001 }, { "canonical": true, "canonical_title": "BLEU: A Method for Automatic Evaluation of Machine Translation", "conflict_flag": false, "editorial": { "authors": "Kishore Papineni et al.", "category": "NLP & Language Models", "confidence": "Medium", "significance": "The metric that made MT progress measurable, virtues and pathologies included.", "venue": "ACL" }, "id": "paper-0046", "language": "en", "level": null, "original_title": null, "work_type": "paper", "year": 2002 }, { "canonical": true, "canonical_title": "Latent Dirichlet Allocation", "conflict_flag": false, "editorial": { "authors": "David M. Blei, Andrew Y. Ng, Michael I. Jordan", "category": "Statistical Learning", "confidence": "High", "significance": "Topic models; a decade of probabilistic text analysis.", "venue": "Journal of Machine Learning Research" }, "id": "paper-0047", "language": "en", "level": null, "original_title": null, "work_type": "paper", "year": 2003 }, { "canonical": true, "canonical_title": "A Neural Probabilistic Language Model", "conflict_flag": false, "editorial": { "authors": "Yoshua Bengio, Rejean Ducharme, Pascal Vincent, Christian Jauvin", "category": "NLP & Language Models", "confidence": "High", "significance": "Word embeddings plus neural language modeling; the LLM lineage starts here.", "venue": "Journal of Machine Learning Research" }, "id": "paper-0048", "language": "en", "level": null, "original_title": null, "work_type": "paper", "year": 2003 }, { "canonical": true, "canonical_title": "A Fast Learning Algorithm for Deep Belief Nets", "conflict_flag": false, "editorial": { "authors": "Geoffrey E. Hinton, Simon Osindero, Yee-Whye Teh", "category": "Neural Networks (Classic)", "confidence": "High", "significance": "Layer-wise pretraining; the paper that relaunched 'deep' learning.", "venue": "Neural Computation" }, "id": "paper-0049", "language": "en", "level": null, "original_title": null, "work_type": "paper", "year": 2006 }, { "canonical": true, "canonical_title": "Reducing the Dimensionality of Data with Neural Networks", "conflict_flag": false, "editorial": { "authors": "Geoffrey E. Hinton, Ruslan Salakhutdinov", "category": "Neural Networks (Classic)", "confidence": "High", "significance": "Deep autoencoders in Science; put deep learning back on the mainstream agenda.", "venue": "Science" }, "id": "paper-0050", "language": "en", "level": null, "original_title": null, "work_type": "paper", "year": 2006 }, { "canonical": true, "canonical_title": "Universal Intelligence: A Definition of Machine Intelligence", "conflict_flag": false, "editorial": { "authors": "Shane Legg, Marcus Hutter", "category": "Meta & Field Essays", "confidence": "Medium", "significance": "The formal definition behind much AGI discourse.", "venue": "Minds and Machines" }, "id": "paper-0051", "language": "en", "level": null, "original_title": null, "work_type": "paper", "year": 2007 }, { "canonical": true, "canonical_title": "The Basic AI Drives", "conflict_flag": false, "editorial": { "authors": "Stephen M. Omohundro", "category": "Safety & Alignment", "confidence": "Medium", "significance": "Instrumental convergence; why capable agents acquire resources and resist shutdown.", "venue": "AGI Conference" }, "id": "paper-0052", "language": "en", "level": null, "original_title": null, "work_type": "paper", "year": 2008 }, { "canonical": true, "canonical_title": "Matrix Factorization Techniques for Recommender Systems", "conflict_flag": false, "editorial": { "authors": "Yehuda Koren, Robert Bell, Chris Volinsky", "category": "Statistical Learning", "confidence": "Medium", "significance": "The Netflix-Prize-era standard for collaborative filtering.", "venue": "IEEE Computer" }, "id": "paper-0053", "language": "en", "level": null, "original_title": null, "work_type": "paper", "year": 2009 }, { "canonical": true, "canonical_title": "ImageNet: A Large-Scale Hierarchical Image Database", "conflict_flag": false, "editorial": { "authors": "Jia Deng, Wei Dong, Richard Socher, Li-Jia Li, Kai Li, Fei-Fei Li", "category": "Deep Learning Era", "confidence": "High", "significance": "The dataset that made the deep learning revolution measurable.", "venue": "CVPR" }, "id": "paper-0054", "language": "en", "level": null, "original_title": null, "work_type": "paper", "year": 2009 }, { "canonical": true, "canonical_title": "The Unreasonable Effectiveness of Data", "conflict_flag": false, "editorial": { "authors": "Alon Halevy, Peter Norvig, Fernando Pereira", "category": "Meta & Field Essays", "confidence": "High", "significance": "Data beats cleverness; the empirical creed of the scaling era, stated early.", "venue": "IEEE Intelligent Systems" }, "id": "paper-0055", "language": "en", "level": null, "original_title": null, "work_type": "paper", "year": 2009 }, { "canonical": true, "canonical_title": "ImageNet Classification with Deep Convolutional Neural Networks", "conflict_flag": false, "editorial": { "authors": "Alex Krizhevsky, Ilya Sutskever, Geoffrey E. Hinton", "category": "Deep Learning Era", "confidence": "High", "significance": "AlexNet; the result that started the modern deep learning era.", "venue": "NeurIPS" }, "id": "paper-0056", "language": "en", "level": null, "original_title": null, "work_type": "paper", "year": 2012 }, { "canonical": true, "canonical_title": "Deep Neural Networks for Acoustic Modeling in Speech Recognition", "conflict_flag": false, "editorial": { "authors": "Geoffrey Hinton et al. (four research groups)", "category": "Deep Learning Era", "confidence": "Medium", "significance": "Deep nets replace decades of speech-recognition engineering.", "venue": "IEEE Signal Processing Magazine" }, "id": "paper-0057", "language": "en", "level": null, "original_title": null, "work_type": "paper", "year": 2012 }, { "canonical": true, "canonical_title": "Fairness Through Awareness", "conflict_flag": false, "editorial": { "authors": "Cynthia Dwork et al.", "category": "Ethics & Fairness", "confidence": "High", "significance": "The formal-fairness research programme begins.", "venue": "ITCS" }, "id": "paper-0058", "language": "en", "level": null, "original_title": null, "work_type": "paper", "year": 2012 }, { "canonical": true, "canonical_title": "A Few Useful Things to Know About Machine Learning", "conflict_flag": false, "editorial": { "authors": "Pedro Domingos", "category": "Meta & Field Essays", "confidence": "High", "significance": "The most-shared practical wisdom paper in ML.", "venue": "Communications of the ACM" }, "id": "paper-0059", "language": "en", "level": null, "original_title": null, "work_type": "paper", "year": 2012 }, { "canonical": true, "canonical_title": "Efficient Estimation of Word Representations in Vector Space", "conflict_flag": false, "editorial": { "authors": "Tomas Mikolov et al.", "category": "NLP & Language Models", "confidence": "High", "significance": "word2vec; cheap, composable word meaning ('king - man + woman').", "venue": "ICLR Workshop" }, "id": "paper-0060", "language": "en", "level": null, "original_title": null, "work_type": "paper", "year": 2013 }, { "canonical": true, "canonical_title": "Dropout: A Simple Way to Prevent Neural Networks from Overfitting", "conflict_flag": false, "editorial": { "authors": "Nitish Srivastava, Geoffrey Hinton, Alex Krizhevsky, Ilya Sutskever, Ruslan Salakhutdinov", "category": "Deep Learning Era", "confidence": "High", "significance": "The defining regularization trick of the era.", "venue": "Journal of Machine Learning Research" }, "id": "paper-0061", "language": "en", "level": null, "original_title": null, "work_type": "paper", "year": 2014 }, { "canonical": true, "canonical_title": "Generative Adversarial Networks", "conflict_flag": false, "editorial": { "authors": "Ian Goodfellow et al.", "category": "Generative Models", "confidence": "High", "significance": "GANs; adversarial training created the modern generative-media era.", "venue": "NeurIPS" }, "id": "paper-0062", "language": "en", "level": null, "original_title": null, "work_type": "paper", "year": 2014 }, { "canonical": true, "canonical_title": "Auto-Encoding Variational Bayes", "conflict_flag": false, "editorial": { "authors": "Diederik P. Kingma, Max Welling", "category": "Generative Models", "confidence": "High", "significance": "VAEs; probabilistic deep generative modeling.", "venue": "ICLR" }, "id": "paper-0063", "language": "en", "level": null, "original_title": null, "work_type": "paper", "year": 2014 }, { "canonical": true, "canonical_title": "Intriguing Properties of Neural Networks", "conflict_flag": false, "editorial": { "authors": "Christian Szegedy et al.", "category": "Security & Privacy", "confidence": "High", "significance": "Discovered adversarial examples; founded ML security.", "venue": "ICLR" }, "id": "paper-0064", "language": "en", "level": null, "original_title": null, "work_type": "paper", "year": 2014 }, { "canonical": true, "canonical_title": "GloVe: Global Vectors for Word Representation", "conflict_flag": false, "editorial": { "authors": "Jeffrey Pennington, Richard Socher, Christopher D. Manning", "category": "NLP & Language Models", "confidence": "Medium", "significance": "The other standard embedding; count-based meets predictive.", "venue": "EMNLP" }, "id": "paper-0065", "language": "en", "level": null, "original_title": null, "work_type": "paper", "year": 2014 }, { "canonical": true, "canonical_title": "Sequence to Sequence Learning with Neural Networks", "conflict_flag": false, "editorial": { "authors": "Ilya Sutskever, Oriol Vinyals, Quoc V. Le", "category": "NLP & Language Models", "confidence": "High", "significance": "Encoder-decoder; end-to-end sequence transduction.", "venue": "NeurIPS" }, "id": "paper-0066", "language": "en", "level": null, "original_title": null, "work_type": "paper", "year": 2014 }, { "canonical": true, "canonical_title": "Very Deep Convolutional Networks for Large-Scale Image Recognition", "conflict_flag": false, "editorial": { "authors": "Karen Simonyan, Andrew Zisserman", "category": "Deep Learning Era", "confidence": "High", "significance": "VGG; depth as the design principle.", "venue": "ICLR" }, "id": "paper-0067", "language": "en", "level": null, "original_title": null, "work_type": "paper", "year": 2015 }, { "canonical": true, "canonical_title": "Going Deeper with Convolutions", "conflict_flag": false, "editorial": { "authors": "Christian Szegedy et al.", "category": "Deep Learning Era", "confidence": "Medium", "significance": "GoogLeNet/Inception; efficient depth via multi-scale modules.", "venue": "CVPR" }, "id": "paper-0068", "language": "en", "level": null, "original_title": null, "work_type": "paper", "year": 2015 }, { "canonical": true, "canonical_title": "Batch Normalization: Accelerating Deep Network Training", "conflict_flag": false, "editorial": { "authors": "Sergey Ioffe, Christian Szegedy", "category": "Deep Learning Era", "confidence": "High", "significance": "Made very deep networks trainable in practice.", "venue": "ICML" }, "id": "paper-0069", "language": "en", "level": null, "original_title": null, "work_type": "paper", "year": 2015 }, { "canonical": true, "canonical_title": "Adam: A Method for Stochastic Optimization", "conflict_flag": false, "editorial": { "authors": "Diederik P. Kingma, Jimmy Ba", "category": "Deep Learning Era", "confidence": "High", "significance": "The default optimizer of deep learning; among the most-cited papers in CS.", "venue": "ICLR" }, "id": "paper-0070", "language": "en", "level": null, "original_title": null, "work_type": "paper", "year": 2015 }, { "canonical": true, "canonical_title": "Deep Learning (review)", "conflict_flag": false, "editorial": { "authors": "Yann LeCun, Yoshua Bengio, Geoffrey Hinton", "category": "Deep Learning Era", "confidence": "High", "significance": "The field's self-definition by its three Turing-Award founders.", "venue": "Nature" }, "id": "paper-0071", "language": "en", "level": null, "original_title": null, "work_type": "paper", "year": 2015 }, { "canonical": true, "canonical_title": "U-Net: Convolutional Networks for Biomedical Image Segmentation", "conflict_flag": false, "editorial": { "authors": "Olaf Ronneberger, Philipp Fischer, Thomas Brox", "category": "Deep Learning Era", "confidence": "Medium", "significance": "The default segmentation architecture, far beyond medicine.", "venue": "MICCAI" }, "id": "paper-0072", "language": "en", "level": null, "original_title": null, "work_type": "paper", "year": 2015 }, { "canonical": true, "canonical_title": "Distilling the Knowledge in a Neural Network", "conflict_flag": false, "editorial": { "authors": "Geoffrey Hinton, Oriol Vinyals, Jeff Dean", "category": "Deep Learning Era", "confidence": "Medium", "significance": "Knowledge distillation; the basis of model compression.", "venue": "NeurIPS Workshop" }, "id": "paper-0073", "language": "en", "level": null, "original_title": null, "work_type": "paper", "year": 2015 }, { "canonical": true, "canonical_title": "Explaining and Harnessing Adversarial Examples", "conflict_flag": false, "editorial": { "authors": "Ian Goodfellow, Jonathon Shlens, Christian Szegedy", "category": "Security & Privacy", "confidence": "Medium", "significance": "FGSM; explained and weaponized adversarial fragility.", "venue": "ICLR" }, "id": "paper-0074", "language": "en", "level": null, "original_title": null, "work_type": "paper", "year": 2015 }, { "canonical": true, "canonical_title": "Neural Machine Translation by Jointly Learning to Align and Translate", "conflict_flag": false, "editorial": { "authors": "Dzmitry Bahdanau, Kyunghyun Cho, Yoshua Bengio", "category": "NLP & Language Models", "confidence": "High", "significance": "Introduced attention; the mechanism that became everything.", "venue": "ICLR" }, "id": "paper-0075", "language": "en", "level": null, "original_title": null, "work_type": "paper", "year": 2015 }, { "canonical": true, "canonical_title": "Human-Level Control Through Deep Reinforcement Learning", "conflict_flag": false, "editorial": { "authors": "Volodymyr Mnih et al.", "category": "Reinforcement Learning", "confidence": "High", "significance": "DQN on Atari; deep RL is born.", "venue": "Nature" }, "id": "paper-0076", "language": "en", "level": null, "original_title": null, "work_type": "paper", "year": 2015 }, { "canonical": true, "canonical_title": "Hidden Technical Debt in Machine Learning Systems", "conflict_flag": false, "editorial": { "authors": "D. Sculley et al.", "category": "Meta & Field Essays", "confidence": "High", "significance": "Why ML systems rot in production; founding text of MLOps.", "venue": "NeurIPS" }, "id": "paper-0077", "language": "en", "level": null, "original_title": null, "work_type": "paper", "year": 2015 }, { "canonical": true, "canonical_title": "XGBoost: A Scalable Tree Boosting System", "conflict_flag": false, "editorial": { "authors": "Tianqi Chen, Carlos Guestrin", "category": "Statistical Learning", "confidence": "High", "significance": "The implementation that made gradient boosting ubiquitous in practice.", "venue": "KDD" }, "id": "paper-0078", "language": "en", "level": null, "original_title": null, "work_type": "paper", "year": 2016 }, { "canonical": true, "canonical_title": "Deep Residual Learning for Image Recognition", "conflict_flag": false, "editorial": { "authors": "Kaiming He, Xiangyu Zhang, Shaoqing Ren, Jian Sun", "category": "Deep Learning Era", "confidence": "High", "significance": "ResNet; skip connections enabled networks of arbitrary depth.", "venue": "CVPR" }, "id": "paper-0079", "language": "en", "level": null, "original_title": null, "work_type": "paper", "year": 2016 }, { "canonical": true, "canonical_title": "You Only Look Once: Unified, Real-Time Object Detection", "conflict_flag": false, "editorial": { "authors": "Joseph Redmon et al.", "category": "Deep Learning Era", "confidence": "Medium", "significance": "YOLO; real-time detection as a single network pass.", "venue": "CVPR" }, "id": "paper-0080", "language": "en", "level": null, "original_title": null, "work_type": "paper", "year": 2016 }, { "canonical": true, "canonical_title": "WaveNet: A Generative Model for Raw Audio", "conflict_flag": false, "editorial": { "authors": "Aaron van den Oord et al.", "category": "Generative Models", "confidence": "Medium", "significance": "Neural audio generation; transformed speech synthesis.", "venue": "arXiv" }, "id": "paper-0081", "language": "en", "level": null, "original_title": null, "work_type": "paper", "year": 2016 }, { "canonical": true, "canonical_title": "Mastering the Game of Go with Deep Neural Networks and Tree Search", "conflict_flag": false, "editorial": { "authors": "David Silver et al.", "category": "Reinforcement Learning", "confidence": "High", "significance": "AlphaGo; the cultural turning point of the deep learning era.", "venue": "Nature" }, "id": "paper-0082", "language": "en", "level": null, "original_title": null, "work_type": "paper", "year": 2016 }, { "canonical": true, "canonical_title": "End-to-End Training of Deep Visuomotor Policies", "conflict_flag": false, "editorial": { "authors": "Sergey Levine et al.", "category": "Robotics", "confidence": "Medium", "significance": "Pixels-to-torques; deep learning enters robotic control.", "venue": "Journal of Machine Learning Research" }, "id": "paper-0083", "language": "en", "level": null, "original_title": null, "work_type": "paper", "year": 2016 }, { "canonical": true, "canonical_title": "Concrete Problems in AI Safety", "conflict_flag": false, "editorial": { "authors": "Dario Amodei et al.", "category": "Safety & Alignment", "confidence": "High", "significance": "Turned AI safety into a concrete ML research agenda.", "venue": "arXiv" }, "id": "paper-0084", "language": "en", "level": null, "original_title": null, "work_type": "paper", "year": 2016 }, { "canonical": true, "canonical_title": "Why Should I Trust You? Explaining the Predictions of Any Classifier", "conflict_flag": false, "editorial": { "authors": "Marco Tulio Ribeiro, Sameer Singh, Carlos Guestrin", "category": "Interpretability", "confidence": "High", "significance": "LIME; model-agnostic local explanation.", "venue": "KDD" }, "id": "paper-0085", "language": "en", "level": null, "original_title": null, "work_type": "paper", "year": 2016 }, { "canonical": true, "canonical_title": "Big Data's Disparate Impact", "conflict_flag": false, "editorial": { "authors": "Solon Barocas, Andrew D. Selbst", "category": "Ethics & Fairness", "confidence": "High", "significance": "The canonical legal analysis of algorithmic discrimination.", "venue": "California Law Review" }, "id": "paper-0086", "language": "en", "level": null, "original_title": null, "work_type": "paper", "year": 2016 }, { "canonical": true, "canonical_title": "Machine Bias", "conflict_flag": false, "editorial": { "authors": "Julia Angwin, Jeff Larson, Surya Mattu, Lauren Kirchner (ProPublica)", "category": "Ethics & Fairness", "confidence": "High", "significance": "The COMPAS investigation; algorithmic injustice becomes front-page news.", "venue": "ProPublica" }, "id": "paper-0087", "language": "en", "level": null, "original_title": null, "work_type": "paper", "year": 2016 }, { "canonical": true, "canonical_title": "Man is to Computer Programmer as Woman is to Homemaker? Debiasing Word Embeddings", "conflict_flag": false, "editorial": { "authors": "Tolga Bolukbasi et al.", "category": "Ethics & Fairness", "confidence": "High", "significance": "Demonstrated social bias embedded in learned representations.", "venue": "NeurIPS" }, "id": "paper-0088", "language": "en", "level": null, "original_title": null, "work_type": "paper", "year": 2016 }, { "canonical": true, "canonical_title": "Equality of Opportunity in Supervised Learning", "conflict_flag": false, "editorial": { "authors": "Moritz Hardt, Eric Price, Nathan Srebro", "category": "Ethics & Fairness", "confidence": "High", "significance": "Defined equalized odds; a standard fairness criterion.", "venue": "NeurIPS" }, "id": "paper-0089", "language": "en", "level": null, "original_title": null, "work_type": "paper", "year": 2016 }, { "canonical": true, "canonical_title": "The Ethics of Algorithms: Mapping the Debate", "conflict_flag": false, "editorial": { "authors": "Brent Mittelstadt et al.", "category": "Ethics & Fairness", "confidence": "Medium", "significance": "The standard map of algorithmic-ethics concerns.", "venue": "Big Data & Society" }, "id": "paper-0090", "language": "en", "level": null, "original_title": null, "work_type": "paper", "year": 2016 }, { "canonical": true, "canonical_title": "Deep Learning with Differential Privacy", "conflict_flag": false, "editorial": { "authors": "Martin Abadi et al.", "category": "Security & Privacy", "confidence": "Medium", "significance": "DP-SGD; private training as a practical method.", "venue": "CCS" }, "id": "paper-0091", "language": "en", "level": null, "original_title": null, "work_type": "paper", "year": 2016 }, { "canonical": true, "canonical_title": "Semi-Supervised Classification with Graph Convolutional Networks", "conflict_flag": false, "editorial": { "authors": "Thomas N. Kipf, Max Welling", "category": "Deep Learning Era", "confidence": "Medium", "significance": "GCNs; the breakthrough that mainstreamed graph deep learning.", "venue": "ICLR" }, "id": "paper-0092", "language": "en", "level": null, "original_title": null, "work_type": "paper", "year": 2017 }, { "canonical": true, "canonical_title": "Attention Is All You Need", "conflict_flag": false, "editorial": { "authors": "Ashish Vaswani et al.", "category": "NLP & Language Models", "confidence": "High", "significance": "The Transformer; the architecture of the modern AI era.", "venue": "NeurIPS" }, "id": "paper-0093", "language": "en", "level": null, "original_title": null, "work_type": "paper", "year": 2017 }, { "canonical": true, "canonical_title": "Mastering the Game of Go Without Human Knowledge", "conflict_flag": false, "editorial": { "authors": "David Silver et al.", "category": "Reinforcement Learning", "confidence": "High", "significance": "AlphaGo Zero; superhuman play from self-play alone.", "venue": "Nature" }, "id": "paper-0094", "language": "en", "level": null, "original_title": null, "work_type": "paper", "year": 2017 }, { "canonical": true, "canonical_title": "Proximal Policy Optimization Algorithms", "conflict_flag": false, "editorial": { "authors": "John Schulman et al.", "category": "Reinforcement Learning", "confidence": "High", "significance": "PPO; the workhorse algorithm, later the engine of RLHF.", "venue": "arXiv" }, "id": "paper-0095", "language": "en", "level": null, "original_title": null, "work_type": "paper", "year": 2017 }, { "canonical": true, "canonical_title": "Deep Reinforcement Learning from Human Preferences", "conflict_flag": false, "editorial": { "authors": "Paul F. Christiano et al.", "category": "Safety & Alignment", "confidence": "High", "significance": "Learning reward from human comparisons; the seed of RLHF.", "venue": "NeurIPS" }, "id": "paper-0096", "language": "en", "level": null, "original_title": null, "work_type": "paper", "year": 2017 }, { "canonical": true, "canonical_title": "A Unified Approach to Interpreting Model Predictions", "conflict_flag": false, "editorial": { "authors": "Scott M. Lundberg, Su-In Lee", "category": "Interpretability", "confidence": "High", "significance": "SHAP; game-theoretic attribution, the industry default.", "venue": "NeurIPS" }, "id": "paper-0097", "language": "en", "level": null, "original_title": null, "work_type": "paper", "year": 2017 }, { "canonical": true, "canonical_title": "Inherent Trade-Offs in the Fair Determination of Risk Scores", "conflict_flag": false, "editorial": { "authors": "Jon Kleinberg, Sendhil Mullainathan, Manish Raghavan", "category": "Ethics & Fairness", "confidence": "High", "significance": "Proved popular fairness definitions are mutually incompatible.", "venue": "ITCS" }, "id": "paper-0098", "language": "en", "level": null, "original_title": null, "work_type": "paper", "year": 2017 }, { "canonical": true, "canonical_title": "Semantics Derived Automatically from Language Corpora Contain Human-Like Biases", "conflict_flag": false, "editorial": { "authors": "Aylin Caliskan, Joanna J. Bryson, Arvind Narayanan", "category": "Ethics & Fairness", "confidence": "High", "significance": "Bias in embeddings, demonstrated with psychometric rigor.", "venue": "Science" }, "id": "paper-0099", "language": "en", "level": null, "original_title": null, "work_type": "paper", "year": 2017 }, { "canonical": true, "canonical_title": "Membership Inference Attacks Against Machine Learning Models", "conflict_flag": false, "editorial": { "authors": "Reza Shokri et al.", "category": "Security & Privacy", "confidence": "Medium", "significance": "Showed models leak whether your data was in the training set.", "venue": "IEEE S&P" }, "id": "paper-0100", "language": "en", "level": null, "original_title": null, "work_type": "paper", "year": 2017 }, { "canonical": true, "canonical_title": "Communication-Efficient Learning of Deep Networks from Decentralized Data", "conflict_flag": false, "editorial": { "authors": "Brendan McMahan et al.", "category": "Security & Privacy", "confidence": "High", "significance": "Federated learning; training without centralizing data.", "venue": "AISTATS" }, "id": "paper-0101", "language": "en", "level": null, "original_title": null, "work_type": "paper", "year": 2017 }, { "canonical": true, "canonical_title": "World Models", "conflict_flag": false, "editorial": { "authors": "David Ha, Juergen Schmidhuber", "category": "Generative Models", "confidence": "Medium", "significance": "Agents learning inside their own learned simulators; ancestor of today's world-model agenda.", "venue": "NeurIPS" }, "id": "paper-0102", "language": "en", "level": null, "original_title": null, "work_type": "paper", "year": 2018 }, { "canonical": true, "canonical_title": "Improving Language Understanding by Generative Pre-Training", "conflict_flag": false, "editorial": { "authors": "Alec Radford, Karthik Narasimhan, Tim Salimans, Ilya Sutskever", "category": "NLP & Language Models", "confidence": "High", "significance": "GPT-1; generative pretraining as the recipe.", "venue": "OpenAI Technical Report" }, "id": "paper-0103", "language": "en", "level": null, "original_title": null, "work_type": "paper", "year": 2018 }, { "canonical": true, "canonical_title": "A General Reinforcement Learning Algorithm That Masters Chess, Shogi, and Go Through Self-Play", "conflict_flag": false, "editorial": { "authors": "David Silver et al.", "category": "Reinforcement Learning", "confidence": "High", "significance": "AlphaZero; one algorithm, three games.", "venue": "Science" }, "id": "paper-0104", "language": "en", "level": null, "original_title": null, "work_type": "paper", "year": 2018 }, { "canonical": true, "canonical_title": "The Mythos of Model Interpretability", "conflict_flag": false, "editorial": { "authors": "Zachary C. Lipton", "category": "Interpretability", "confidence": "Medium", "significance": "Disciplined the field's vocabulary about what 'interpretable' means.", "venue": "ACM Queue / CACM" }, "id": "paper-0105", "language": "en", "level": null, "original_title": null, "work_type": "paper", "year": 2018 }, { "canonical": true, "canonical_title": "Gender Shades: Intersectional Accuracy Disparities in Commercial Gender Classification", "conflict_flag": false, "editorial": { "authors": "Joy Buolamwini, Timnit Gebru", "category": "Ethics & Fairness", "confidence": "High", "significance": "Audit that changed commercial face-analysis products and founded audit culture.", "venue": "FAT*" }, "id": "paper-0106", "language": "en", "level": null, "original_title": null, "work_type": "paper", "year": 2018 }, { "canonical": true, "canonical_title": "Counterfactual Explanations Without Opening the Black Box", "conflict_flag": false, "editorial": { "authors": "Sandra Wachter, Brent Mittelstadt, Chris Russell", "category": "Governance & Policy", "confidence": "Medium", "significance": "Linked explanation methods to GDPR; the legal-technical bridge.", "venue": "Harvard Journal of Law & Technology" }, "id": "paper-0107", "language": "en", "level": null, "original_title": null, "work_type": "paper", "year": 2018 }, { "canonical": true, "canonical_title": "The Malicious Use of Artificial Intelligence: Forecasting, Prevention, and Mitigation", "conflict_flag": false, "editorial": { "authors": "Miles Brundage et al.", "category": "Governance & Policy", "confidence": "High", "significance": "The first broad misuse-threat assessment across digital, physical, political security.", "venue": "arXiv" }, "id": "paper-0108", "language": "en", "level": null, "original_title": null, "work_type": "paper", "year": 2018 }, { "canonical": true, "canonical_title": "BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding", "conflict_flag": false, "editorial": { "authors": "Jacob Devlin, Ming-Wei Chang, Kenton Lee, Kristina Toutanova", "category": "NLP & Language Models", "confidence": "High", "significance": "Pretrain-finetune became the default NLP paradigm.", "venue": "NAACL" }, "id": "paper-0109", "language": "en", "level": null, "original_title": null, "work_type": "paper", "year": 2019 }, { "canonical": true, "canonical_title": "Language Models are Unsupervised Multitask Learners", "conflict_flag": false, "editorial": { "authors": "Alec Radford et al.", "category": "NLP & Language Models", "confidence": "High", "significance": "GPT-2; scale yields zero-shot task behaviour, and the first staged-release debate.", "venue": "OpenAI Technical Report" }, "id": "paper-0110", "language": "en", "level": null, "original_title": null, "work_type": "paper", "year": 2019 }, { "canonical": true, "canonical_title": "Grandmaster Level in StarCraft II Using Multi-Agent Reinforcement Learning", "conflict_flag": false, "editorial": { "authors": "Oriol Vinyals et al.", "category": "Reinforcement Learning", "confidence": "Medium", "significance": "AlphaStar; RL in a real-time, partially observed strategy game.", "venue": "Nature" }, "id": "paper-0111", "language": "en", "level": null, "original_title": null, "work_type": "paper", "year": 2019 }, { "canonical": true, "canonical_title": "Risks from Learned Optimization in Advanced Machine Learning Systems", "conflict_flag": false, "editorial": { "authors": "Evan Hubinger et al.", "category": "Safety & Alignment", "confidence": "Medium", "significance": "Mesa-optimization and deceptive alignment; core inner-alignment concepts.", "venue": "arXiv" }, "id": "paper-0112", "language": "en", "level": null, "original_title": null, "work_type": "paper", "year": 2019 }, { "canonical": true, "canonical_title": "Stop Explaining Black Box Machine Learning Models for High Stakes Decisions", "conflict_flag": false, "editorial": { "authors": "Cynthia Rudin", "category": "Interpretability", "confidence": "High", "significance": "Argues for inherently interpretable models where stakes are high.", "venue": "Nature Machine Intelligence" }, "id": "paper-0113", "language": "en", "level": null, "original_title": null, "work_type": "paper", "year": 2019 }, { "canonical": true, "canonical_title": "Model Cards for Model Reporting", "conflict_flag": false, "editorial": { "authors": "Margaret Mitchell et al.", "category": "Ethics & Fairness", "confidence": "High", "significance": "The documentation standard for released models.", "venue": "FAT*" }, "id": "paper-0114", "language": "en", "level": null, "original_title": null, "work_type": "paper", "year": 2019 }, { "canonical": true, "canonical_title": "Dissecting Racial Bias in an Algorithm Used to Manage the Health of Populations", "conflict_flag": false, "editorial": { "authors": "Ziad Obermeyer et al.", "category": "Ethics & Fairness", "confidence": "High", "significance": "Showed a deployed health algorithm systematically disadvantaged Black patients.", "venue": "Science" }, "id": "paper-0115", "language": "en", "level": null, "original_title": null, "work_type": "paper", "year": 2019 }, { "canonical": true, "canonical_title": "Energy and Policy Considerations for Deep Learning in NLP", "conflict_flag": false, "editorial": { "authors": "Emma Strubell, Ananya Ganesh, Andrew McCallum", "category": "Ethics & Fairness", "confidence": "High", "significance": "Put training cost and carbon on the research agenda.", "venue": "ACL" }, "id": "paper-0116", "language": "en", "level": null, "original_title": null, "work_type": "paper", "year": 2019 }, { "canonical": true, "canonical_title": "The Global Landscape of AI Ethics Guidelines", "conflict_flag": false, "editorial": { "authors": "Anna Jobin, Marcello Ienca, Effy Vayena", "category": "Governance & Policy", "confidence": "High", "significance": "Mapped 84 guidelines; showed convergence on principles and divergence on practice.", "venue": "Nature Machine Intelligence" }, "id": "paper-0117", "language": "en", "level": null, "original_title": null, "work_type": "paper", "year": 2019 }, { "canonical": true, "canonical_title": "The Bitter Lesson", "conflict_flag": false, "editorial": { "authors": "Richard Sutton", "category": "Meta & Field Essays", "confidence": "High", "significance": "Compute-leveraging general methods beat human-knowledge engineering; the era's most-quoted essay.", "venue": "Essay (incompleteideas.net)" }, "id": "paper-0118", "language": "en", "level": null, "original_title": null, "work_type": "paper", "year": 2019 }, { "canonical": true, "canonical_title": "On the Measure of Intelligence", "conflict_flag": false, "editorial": { "authors": "Francois Chollet", "category": "Meta & Field Essays", "confidence": "Medium", "significance": "Skill-acquisition efficiency as the definition of intelligence; basis of the ARC benchmark.", "venue": "arXiv" }, "id": "paper-0119", "language": "en", "level": null, "original_title": null, "work_type": "paper", "year": 2019 }, { "canonical": true, "canonical_title": "Denoising Diffusion Probabilistic Models", "conflict_flag": false, "editorial": { "authors": "Jonathan Ho, Ajay Jain, Pieter Abbeel", "category": "Generative Models", "confidence": "High", "significance": "Made diffusion the dominant image-generation paradigm.", "venue": "NeurIPS" }, "id": "paper-0120", "language": "en", "level": null, "original_title": null, "work_type": "paper", "year": 2020 }, { "canonical": true, "canonical_title": "Neural Radiance Fields (NeRF): Representing Scenes for View Synthesis", "conflict_flag": false, "editorial": { "authors": "Ben Mildenhall et al.", "category": "Generative Models", "confidence": "Medium", "significance": "Learned 3D scene representation; new field of neural rendering.", "venue": "ECCV" }, "id": "paper-0121", "language": "en", "level": null, "original_title": null, "work_type": "paper", "year": 2020 }, { "canonical": true, "canonical_title": "Exploring the Limits of Transfer Learning with a Unified Text-to-Text Transformer", "conflict_flag": false, "editorial": { "authors": "Colin Raffel et al.", "category": "NLP & Language Models", "confidence": "Medium", "significance": "T5; everything is text-to-text.", "venue": "Journal of Machine Learning Research" }, "id": "paper-0122", "language": "en", "level": null, "original_title": null, "work_type": "paper", "year": 2020 }, { "canonical": true, "canonical_title": "Language Models are Few-Shot Learners", "conflict_flag": false, "editorial": { "authors": "Tom B. Brown et al.", "category": "NLP & Language Models", "confidence": "High", "significance": "GPT-3; in-context learning and the scaling thesis made undeniable.", "venue": "NeurIPS" }, "id": "paper-0123", "language": "en", "level": null, "original_title": null, "work_type": "paper", "year": 2020 }, { "canonical": true, "canonical_title": "Scaling Laws for Neural Language Models", "conflict_flag": false, "editorial": { "authors": "Jared Kaplan et al.", "category": "Scaling & Frontier", "confidence": "High", "significance": "Loss as a power law of compute, data, parameters; the industry's planning document.", "venue": "arXiv" }, "id": "paper-0124", "language": "en", "level": null, "original_title": null, "work_type": "paper", "year": 2020 }, { "canonical": true, "canonical_title": "Retrieval-Augmented Generation for Knowledge-Intensive NLP Tasks", "conflict_flag": false, "editorial": { "authors": "Patrick Lewis et al.", "category": "NLP & Language Models", "confidence": "High", "significance": "RAG; grounding generation in retrieved evidence.", "venue": "NeurIPS" }, "id": "paper-0125", "language": "en", "level": null, "original_title": null, "work_type": "paper", "year": 2020 }, { "canonical": true, "canonical_title": "Mastering Atari, Go, Chess and Shogi by Planning with a Learned Model", "conflict_flag": false, "editorial": { "authors": "Julian Schrittwieser et al.", "category": "Reinforcement Learning", "confidence": "Medium", "significance": "MuZero; planning without knowing the rules.", "venue": "Nature" }, "id": "paper-0126", "language": "en", "level": null, "original_title": null, "work_type": "paper", "year": 2020 }, { "canonical": true, "canonical_title": "Zoom In: An Introduction to Circuits", "conflict_flag": false, "editorial": { "authors": "Chris Olah et al.", "category": "Interpretability", "confidence": "Medium", "significance": "Founded mechanistic interpretability: studying networks like organisms.", "venue": "Distill" }, "id": "paper-0127", "language": "en", "level": null, "original_title": null, "work_type": "paper", "year": 2020 }, { "canonical": true, "canonical_title": "Closing the AI Accountability Gap: Defining an End-to-End Framework for Internal Algorithmic Auditing", "conflict_flag": false, "editorial": { "authors": "Inioluwa Deborah Raji et al.", "category": "Governance & Policy", "confidence": "High", "significance": "The reference framework for internal AI audit practice.", "venue": "FAT*" }, "id": "paper-0128", "language": "en", "level": null, "original_title": null, "work_type": "paper", "year": 2020 }, { "canonical": true, "canonical_title": "An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale", "conflict_flag": false, "editorial": { "authors": "Alexey Dosovitskiy et al.", "category": "Deep Learning Era", "confidence": "High", "significance": "ViT; transformers displace convolutions in vision.", "venue": "ICLR" }, "id": "paper-0129", "language": "en", "level": null, "original_title": null, "work_type": "paper", "year": 2021 }, { "canonical": true, "canonical_title": "Learning Transferable Visual Models From Natural Language Supervision", "conflict_flag": false, "editorial": { "authors": "Alec Radford et al.", "category": "Deep Learning Era", "confidence": "High", "significance": "CLIP; vision-language alignment underpinning multimodal AI.", "venue": "ICML" }, "id": "paper-0130", "language": "en", "level": null, "original_title": null, "work_type": "paper", "year": 2021 }, { "canonical": true, "canonical_title": "Evaluating Large Language Models Trained on Code", "conflict_flag": false, "editorial": { "authors": "Mark Chen et al.", "category": "LLM Methods", "confidence": "Medium", "significance": "Codex; LLMs write code, the capability that transformed software work.", "venue": "arXiv" }, "id": "paper-0131", "language": "en", "level": null, "original_title": null, "work_type": "paper", "year": 2021 }, { "canonical": true, "canonical_title": "Unsolved Problems in ML Safety", "conflict_flag": false, "editorial": { "authors": "Dan Hendrycks et al.", "category": "Safety & Alignment", "confidence": "Medium", "significance": "The mainstream-ML framing of robustness, monitoring, alignment, systemic safety.", "venue": "arXiv" }, "id": "paper-0132", "language": "en", "level": null, "original_title": null, "work_type": "paper", "year": 2021 }, { "canonical": true, "canonical_title": "Datasheets for Datasets", "conflict_flag": false, "editorial": { "authors": "Timnit Gebru et al.", "category": "Ethics & Fairness", "confidence": "High", "significance": "Provenance documentation for training data.", "venue": "Communications of the ACM" }, "id": "paper-0133", "language": "en", "level": null, "original_title": null, "work_type": "paper", "year": 2021 }, { "canonical": true, "canonical_title": "On the Dangers of Stochastic Parrots: Can Language Models Be Too Big?", "conflict_flag": false, "editorial": { "authors": "Emily M. Bender, Timnit Gebru, Angelina McMillan-Major, Margaret Mitchell", "category": "Ethics & Fairness", "confidence": "High", "significance": "The defining critique of the LLM paradigm, and the paper behind Google's Gebru affair.", "venue": "FAccT" }, "id": "paper-0134", "language": "en", "level": null, "original_title": null, "work_type": "paper", "year": 2021 }, { "canonical": true, "canonical_title": "Extracting Training Data from Large Language Models", "conflict_flag": false, "editorial": { "authors": "Nicholas Carlini et al.", "category": "Security & Privacy", "confidence": "Medium", "significance": "LLMs memorize and can regurgitate training data.", "venue": "USENIX Security" }, "id": "paper-0135", "language": "en", "level": null, "original_title": null, "work_type": "paper", "year": 2021 }, { "canonical": true, "canonical_title": "Highly Accurate Protein Structure Prediction with AlphaFold", "conflict_flag": false, "editorial": { "authors": "John Jumper et al.", "category": "AI for Science", "confidence": "High", "significance": "Solved a fifty-year grand challenge; the Nobel-recognized proof of AI for science.", "venue": "Nature" }, "id": "paper-0136", "language": "en", "level": null, "original_title": null, "work_type": "paper", "year": 2021 }, { "canonical": true, "canonical_title": "The Hardware Lottery", "conflict_flag": false, "editorial": { "authors": "Sara Hooker", "category": "Meta & Field Essays", "confidence": "Medium", "significance": "Research directions win because hardware favors them; a structural critique of progress.", "venue": "Communications of the ACM" }, "id": "paper-0137", "language": "en", "level": null, "original_title": null, "work_type": "paper", "year": 2021 }, { "canonical": true, "canonical_title": "High-Resolution Image Synthesis with Latent Diffusion Models", "conflict_flag": false, "editorial": { "authors": "Robin Rombach et al.", "category": "Generative Models", "confidence": "High", "significance": "Stable Diffusion; open-weights image generation at consumer scale.", "venue": "CVPR" }, "id": "paper-0138", "language": "en", "level": null, "original_title": null, "work_type": "paper", "year": 2022 }, { "canonical": true, "canonical_title": "Training Compute-Optimal Large Language Models", "conflict_flag": false, "editorial": { "authors": "Jordan Hoffmann et al.", "category": "Scaling & Frontier", "confidence": "High", "significance": "Chinchilla; rebalanced the field toward data-optimal training.", "venue": "NeurIPS" }, "id": "paper-0139", "language": "en", "level": null, "original_title": null, "work_type": "paper", "year": 2022 }, { "canonical": true, "canonical_title": "LoRA: Low-Rank Adaptation of Large Language Models", "conflict_flag": false, "editorial": { "authors": "Edward J. Hu et al.", "category": "LLM Methods", "confidence": "High", "significance": "Made finetuning large models affordable; standard adaptation method.", "venue": "ICLR" }, "id": "paper-0140", "language": "en", "level": null, "original_title": null, "work_type": "paper", "year": 2022 }, { "canonical": true, "canonical_title": "Switch Transformers: Scaling to Trillion Parameter Models with Simple and Efficient Sparsity", "conflict_flag": false, "editorial": { "authors": "William Fedus, Barret Zoph, Noam Shazeer", "category": "Scaling & Frontier", "confidence": "Medium", "significance": "Mixture-of-experts at scale; the sparse path to frontier capability.", "venue": "Journal of Machine Learning Research" }, "id": "paper-0141", "language": "en", "level": null, "original_title": null, "work_type": "paper", "year": 2022 }, { "canonical": true, "canonical_title": "Chain-of-Thought Prompting Elicits Reasoning in Large Language Models", "conflict_flag": false, "editorial": { "authors": "Jason Wei et al.", "category": "LLM Methods", "confidence": "High", "significance": "Stepwise prompting unlocks latent reasoning; opened the reasoning-model agenda.", "venue": "NeurIPS" }, "id": "paper-0142", "language": "en", "level": null, "original_title": null, "work_type": "paper", "year": 2022 }, { "canonical": true, "canonical_title": "Emergent Abilities of Large Language Models", "conflict_flag": false, "editorial": { "authors": "Jason Wei et al.", "category": "Scaling & Frontier", "confidence": "Medium", "significance": "Named (and contested) the phenomenon of capability jumps with scale.", "venue": "TMLR" }, "id": "paper-0143", "language": "en", "level": null, "original_title": null, "work_type": "paper", "year": 2022 }, { "canonical": true, "canonical_title": "Training Language Models to Follow Instructions with Human Feedback", "conflict_flag": false, "editorial": { "authors": "Long Ouyang et al.", "category": "Safety & Alignment", "confidence": "High", "significance": "InstructGPT; RLHF at scale, the technique behind aligned chat models.", "venue": "NeurIPS" }, "id": "paper-0144", "language": "en", "level": null, "original_title": null, "work_type": "paper", "year": 2022 }, { "canonical": true, "canonical_title": "Constitutional AI: Harmlessness from AI Feedback", "conflict_flag": false, "editorial": { "authors": "Yuntao Bai et al.", "category": "Safety & Alignment", "confidence": "High", "significance": "Alignment via explicit principles and AI feedback; reduced dependence on human labeling.", "venue": "arXiv" }, "id": "paper-0145", "language": "en", "level": null, "original_title": null, "work_type": "paper", "year": 2022 }, { "canonical": true, "canonical_title": "Toy Models of Superposition", "conflict_flag": false, "editorial": { "authors": "Nelson Elhage et al.", "category": "Interpretability", "confidence": "Medium", "significance": "Why features share neurons; the core obstacle to reading networks.", "venue": "Anthropic / Transformer Circuits" }, "id": "paper-0146", "language": "en", "level": null, "original_title": null, "work_type": "paper", "year": 2022 }, { "canonical": true, "canonical_title": "Discovering Faster Matrix Multiplication Algorithms with Reinforcement Learning", "conflict_flag": false, "editorial": { "authors": "Alhussein Fawzi et al.", "category": "AI for Science", "confidence": "Medium", "significance": "AlphaTensor; AI finds new mathematics.", "venue": "Nature" }, "id": "paper-0147", "language": "en", "level": null, "original_title": null, "work_type": "paper", "year": 2022 }, { "canonical": true, "canonical_title": "Segment Anything", "conflict_flag": false, "editorial": { "authors": "Alexander Kirillov et al.", "category": "Deep Learning Era", "confidence": "Medium", "significance": "Promptable foundation model for segmentation.", "venue": "ICCV" }, "id": "paper-0148", "language": "en", "level": null, "original_title": null, "work_type": "paper", "year": 2023 }, { "canonical": true, "canonical_title": "ReAct: Synergizing Reasoning and Acting in Language Models", "conflict_flag": false, "editorial": { "authors": "Shunyu Yao et al.", "category": "LLM Methods", "confidence": "Medium", "significance": "Reason-act loops; the template for LLM agents.", "venue": "ICLR" }, "id": "paper-0149", "language": "en", "level": null, "original_title": null, "work_type": "paper", "year": 2023 }, { "canonical": true, "canonical_title": "Toolformer: Language Models Can Teach Themselves to Use Tools", "conflict_flag": false, "editorial": { "authors": "Timo Schick et al.", "category": "LLM Methods", "confidence": "Medium", "significance": "Self-supervised tool use; agents calling APIs.", "venue": "NeurIPS" }, "id": "paper-0150", "language": "en", "level": null, "original_title": null, "work_type": "paper", "year": 2023 }, { "canonical": true, "canonical_title": "LLaMA: Open and Efficient Foundation Language Models", "conflict_flag": false, "editorial": { "authors": "Hugo Touvron et al.", "category": "Scaling & Frontier", "confidence": "High", "significance": "The open-weights line that created today's open-model ecosystem.", "venue": "arXiv" }, "id": "paper-0151", "language": "en", "level": null, "original_title": null, "work_type": "paper", "year": 2023 }, { "canonical": true, "canonical_title": "GPT-4 Technical Report", "conflict_flag": false, "editorial": { "authors": "OpenAI", "category": "Scaling & Frontier", "confidence": "High", "significance": "The frontier capability report; also the moment training details went dark.", "venue": "arXiv" }, "id": "paper-0152", "language": "en", "level": null, "original_title": null, "work_type": "paper", "year": 2023 }, { "canonical": true, "canonical_title": "Sparks of Artificial General Intelligence: Early Experiments with GPT-4", "conflict_flag": false, "editorial": { "authors": "Sebastien Bubeck et al.", "category": "Scaling & Frontier", "confidence": "High", "significance": "The most-debated capability claim of the era; framed the AGI-proximity argument.", "venue": "arXiv" }, "id": "paper-0153", "language": "en", "level": null, "original_title": null, "work_type": "paper", "year": 2023 }, { "canonical": true, "canonical_title": "Mamba: Linear-Time Sequence Modeling with Selective State Spaces", "conflict_flag": false, "editorial": { "authors": "Albert Gu, Tri Dao", "category": "LLM Methods", "confidence": "Medium", "significance": "The leading post-Transformer architecture candidate.", "venue": "arXiv" }, "id": "paper-0154", "language": "en", "level": null, "original_title": null, "work_type": "paper", "year": 2023 }, { "canonical": true, "canonical_title": "Universal and Transferable Adversarial Attacks on Aligned Language Models", "conflict_flag": false, "editorial": { "authors": "Andy Zou et al.", "category": "Security & Privacy", "confidence": "Medium", "significance": "Automated jailbreaks; alignment as an attack surface.", "venue": "arXiv" }, "id": "paper-0155", "language": "en", "level": null, "original_title": null, "work_type": "paper", "year": 2023 }, { "canonical": true, "canonical_title": "Towards Monosemanticity: Decomposing Language Models with Dictionary Learning", "conflict_flag": false, "editorial": { "authors": "Trenton Bricken et al.", "category": "Interpretability", "confidence": "Medium", "significance": "Sparse autoencoders extract human-legible features from LLMs.", "venue": "Anthropic / Transformer Circuits" }, "id": "paper-0156", "language": "en", "level": null, "original_title": null, "work_type": "paper", "year": 2023 }, { "canonical": true, "canonical_title": "Frontier AI Regulation: Managing Emerging Risks to Public Safety", "conflict_flag": false, "editorial": { "authors": "Markus Anderljung et al.", "category": "Governance & Policy", "confidence": "Medium", "significance": "The reference proposal for frontier-model regulatory architecture.", "venue": "arXiv" }, "id": "paper-0157", "language": "en", "level": null, "original_title": null, "work_type": "paper", "year": 2023 }, { "canonical": true, "canonical_title": "GPTs are GPTs: An Early Look at the Labor Market Impact Potential of Large Language Models", "conflict_flag": false, "editorial": { "authors": "Tyna Eloundou, Sam Manning, Pamela Mishkin, Daniel Rock", "category": "Economics of AI", "confidence": "Medium", "significance": "First systematic occupational-exposure estimates for LLMs.", "venue": "arXiv (later Science)" }, "id": "paper-0158", "language": "en", "level": null, "original_title": null, "work_type": "paper", "year": 2023 }, { "canonical": true, "canonical_title": "Sleeper Agents: Training Deceptive LLMs That Persist Through Safety Training", "conflict_flag": false, "editorial": { "authors": "Evan Hubinger et al.", "category": "Safety & Alignment", "confidence": "Medium", "significance": "Empirical evidence that deceptive behaviour can survive standard safety training.", "venue": "arXiv" }, "id": "paper-0159", "language": "en", "level": null, "original_title": null, "work_type": "paper", "year": 2024 }, { "canonical": true, "canonical_title": "Managing Extreme AI Risks Amid Rapid Progress", "conflict_flag": false, "editorial": { "authors": "Yoshua Bengio et al.", "category": "Safety & Alignment", "confidence": "High", "significance": "Consensus statement by senior researchers on frontier-risk preparedness.", "venue": "Science" }, "id": "paper-0160", "language": "en", "level": null, "original_title": null, "work_type": "paper", "year": 2024 }, { "canonical": true, "canonical_title": "Accurate Structure Prediction of Biomolecular Interactions with AlphaFold 3", "conflict_flag": false, "editorial": { "authors": "Josh Abramson et al.", "category": "AI for Science", "confidence": "Medium", "significance": "Extended structure prediction to complexes and drug-relevant interactions.", "venue": "Nature" }, "id": "paper-0161", "language": "en", "level": null, "original_title": null, "work_type": "paper", "year": 2024 }, { "canonical": true, "canonical_title": "DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learning", "conflict_flag": false, "editorial": { "authors": "DeepSeek-AI", "category": "Scaling & Frontier", "confidence": "High", "significance": "Open reasoning model from China that reset assumptions about cost and access.", "venue": "arXiv" }, "id": "paper-0162", "language": "en", "level": null, "original_title": null, "work_type": "paper", "year": 2025 } ]