Research Hall of Fame
Important papers from the history of AI, computing, and cognitive science, explained in plain English.
Browse topics All topics
Agents
Building Autonomous Agents for Minecraft
Voyager is an embodied agent that uses GPT-4 to autonomously learn and store executable code skills for complex task completion in Minecraft.
2023Building Autonomous Agents with Social Behavior
The researchers developed an agent architecture using memory, reflection, and planning modules to simulate believable social interactions between independent software agents.
2023Teaching Language Models to Use APIs
The researchers developed a self-supervised method allowing language models to learn how and when to invoke external APIs to overcome limitations in calculation and factual retrieval.
2022Interleaving Reasoning and Acting in LLMs
The ReAct paradigm enables language models to solve complex tasks by interleaving verbal reasoning with external API actions to dynamically update plans.
Artificial Intelligence
Predicting Protein Structures With AlphaFold
The paper introduces AlphaFold, an entirely redesigned neural network model that predicts three-dimensional protein structures from amino acid sequences with high accuracy.
2020GPT-3 Few-Shot Language Learning Performance
The paper introduces a 175-billion parameter model capable of performing tasks with zero or few examples provided in the prompt without needing model weight updates.
2018Training Deep Bidirectional Language Models
This paper introduces BERT to enable pre-training of deep bidirectional representations by jointly conditioning on both left and right context in all layers.
2017Replacing Recurrence with Parallel Attention Mechanisms
The Transformer architecture replaces sequential neural networks with an attention based mechanism that enables full parallelization during training.
2014Adam: A New Stochastic Optimization Method
The paper introduces Adam, an algorithm for first-order gradient-based optimization designed for large-scale data and parameter sets.
2014Adversarial Training for Generative Models
The paper introduces a framework where two neural networks compete against each other to generate realistic data without requiring Markov chains or complex probabilistic inference.
2013Efficiently Creating High Quality Word Representations
The paper introduces two novel neural network architectures designed to learn word representations more efficiently at scale.
1997Long Short Term Memory Networks
The paper introduces a recurrent neural network architecture called long short term memory to store information over extended time intervals by preventing decaying error backflow.
Benchmarks & Evals
Evaluating Language Models on Real Software Issues
The paper introduces SWE-bench, a new benchmark that tests language models on resolving real-world GitHub issues by navigating large repositories and executing unit tests.
2021Evaluating Large Language Models Trained on Code
The paper investigates the program synthesis and code-writing capabilities of large language models trained on code, focusing on generating standalone Python functions from docstrings and measuring functional correctness.
2020Measuring Multitask Language Understanding Across Subjects
The authors designed the Massive Multitask Language Understanding benchmark covering 57 subjects to evaluate how well language models learn and apply knowledge across multiple domains.
2018A Unified Benchmark for Natural Language Understanding
The paper introduces the General Language Understanding Evaluation benchmark suite to test natural language processing models across diverse tasks without relying on task-specific designs.
2009Building the ImageNet Large Scale Database
The paper presents ImageNet, a large scale hierarchical image database built upon the WordNet structure to help solve the critical problem of harnessing massive Internet image data.
Computer Vision
Segment Anything Model for Image Segmentation
The paper introduces a foundation model and data engine for promptable image segmentation that enables zero shot transfer to downstream tasks.
2021High Resolution Image Synthesis with Latent Diffusion Models
The paper introduces Latent Diffusion Models, which apply diffusion models in a lower dimensional latent space using pretrained autoencoders to reduce computational costs while achieving state-of-the-art image synthesis performance.
2020Using Transformers for Image Recognition
The paper demonstrates that a pure Transformer architecture, applied directly to sequences of image patches, can achieve excellent image classification results compared to state-of-the-art convolutional networks while requiring substantially fewer computational resources to train.
2020Generating High Quality Images with Diffusion
This paper introduces a diffusion probabilistic model capable of generating high quality image samples that compete with other deep generative models.
2015Training Deeper Neural Networks with Residuals
The paper introduces residual learning and shortcut connections to successfully train substantially deeper neural networks without suffering from degradation and vanishing gradients.
2015Real Time Object Detection With A Single Network
This paper reframes object detection as a single regression problem, allowing a single neural network to predict bounding boxes and classes directly from full images in one evaluation.
2015U-Net for Biomedical Image Segmentation
The paper introduces U-Net, a convolutional network architecture designed for precise biomedical image segmentation using very few training images.
2012Classifying Millions of Images with Deep Networks
The paper uses a large, deep convolutional neural network to classify 1.2 million high-resolution images into 1000 different classes.
Efficiency & Inference
Accelerating Transformer Inference Through Speculative Decoding
The paper introduces speculative decoding, a method to speed up inference in large autoregressive models by using a smaller approximation model to guess tokens in parallel before evaluating them with the target model.
20228-bit Matrix Multiplication for Large Language Models
The paper introduces LLM.int8(), a quantization method that enables zero-degradation inference for large language models up to 175B parameters by isolating emergent outlier features into a 16-bit matrix multiplication.
2022Fast and Memory Efficient Exact Attention
The authors introduce FlashAttention, a restructuring of the self-attention mechanism that is fast and memory efficient for long sequences.
2018Finding Trainable Subnetworks in Random Neural Networks
This paper investigates whether dense, randomly initialized neural networks contain sparse subnetworks that can be trained in isolation to match the original network's accuracy.
2017Scaling Neural Networks with Mixture of Experts
The paper introduces the Sparsely-Gated Mixture-of-Experts Layer, which dramatically increases network capacity and performance while maintaining computational efficiency.
2015Distilling Knowledge Into Smaller Models
The paper introduces knowledge distillation, a technique to transfer generalization ability from cumbersome ensembles or large neural networks into smaller deployment-ready models.
Multimodal
Building a General Purpose Visual Assistant
The paper introduces LLaVA, a multimodal visual assistant built by connecting a visual encoder to a language model and fine-tuning them on automatically generated instruction data.
2022Robust Speech Recognition via Large Scale Weak Supervision
The paper develops a robust speech recognition system that works reliably out of the box across a broad range of environments, tasks, and languages without requiring supervised fine tuning of a decoder for every deployment distribution.
2022Flamingo Visual Language Model
Flamingo is a visual language model that adapts to novel multimodal tasks using only a handful of annotated examples.
2021Learning Visual Models from Language Supervision
This paper introduces a method called CLIP, which trains computer vision models to understand images by associating them with natural language captions, significantly improving zero-shot performance without needing task-specific labeled data.
2021Zero Shot Text to Image Generation
The paper introduces a method that generates images directly from text by compressing images into discrete tokens and training a large sparse transformer to model text and image tokens together.
Reasoning
Incentivizing LLM Reasoning via Pure Reinforcement Learning
The paper introduces DeepSeek-R1, demonstrating that large language models can develop advanced reasoning capabilities through pure reinforcement learning without requiring prior supervised fine-tuning.
2023Verifying Step by Step Reasoning in LLMs
The paper compares outcome supervision with process supervision for training large language models to reduce logical mistakes and hallucinations during complex multi-step reasoning tasks.
2023Tree Search for Language Models
The paper introduces a framework called Tree of Thoughts that frames problem-solving as a search over a tree of partial solutions, allowing language models to perform deliberate exploration, lookahead, and backtracking.
2022Improving Language Model Reasoning with Self-Consistency
The paper introduces self-consistency, a decoding strategy that replaces greedy decoding with sampling and majority voting to significantly improve reasoning accuracy in language models.
2022Teaching Large Language Models To Reason
The paper introduces chain-of-thought prompting, a technique where models generate intermediate reasoning steps before answering, which significantly improves their performance on complex reasoning tasks.
Reinforcement Learning
Better Update Methods for Reinforcement Learning
The authors propose a novel surrogate objective function that enables multiple epochs of minibatch updates for proximal policy optimization, improving efficiency and stability.
2016Mastering Go With Deep Neural Networks
The paper introduces AlphaGo, a system that combines deep neural networks and tree search to master the game of Go.
2013Playing Atari Games Using Deep Reinforcement Learning
The paper presents a deep learning model that successfully learns control policies directly from high-dimensional raw video data using reinforcement learning.
Robotics
Connecting Vision Language Models To Robots
The paper introduces vision language action models, which incorporate internet scale web data directly into robotic control to improve generalization and semantic reasoning.
2022Robotics Transformer for Real-World Control
The paper introduces RT-1, a large multi-task backbone model built on robotic task data to achieve zero-shot generalization to new tasks, environments, and objects.
2015End-to-End Training for Robotic Control
The paper investigates whether training perception and control systems jointly end-to-end yields better performance than training each component separately for autonomous robotic operation.
Safety & Alignment
Constitutional AI for Harmless Assistants
The paper introduces Constitutional AI to train helpful and harmless AI assistants using self-critique and AI feedback instead of relying solely on manual human oversight.
2022Training Language Models With Human Feedback
The paper introduces a method to align large language models with user intent using supervised learning and reinforcement learning from human feedback.
2017Teaching AI With Human Preferences
The paper introduces a method to train reinforcement learning systems on complex tasks using human feedback instead of well-specified reward functions.
2016Avoiding Accidents in Machine Learning Systems
The paper identifies and analyzes five concrete problems related to accident risks in machine learning systems.
Training & Fine-Tuning
Simpler Control of Language Models
Researchers developed a simpler method to control language model behavior directly from human preferences without training a separate reward model or using reinforcement learning loops.
2022Finding the Optimal Size for Language Models
The paper investigates the optimal model size and number of tokens for training a transformer language model under a given compute budget, finding that current large language models are significantly undertrained.
2021Efficient Fine-Tuning for Large Language Models
The paper introduces Low-Rank Adaptation, a method that freezes pre-trained model weights and trains low-rank decomposition matrices to update models efficiently without introducing inference latency.
2018Universal Language Model Fine-Tuning
The paper introduces ULMFiT, a universal transfer learning method that significantly outperforms existing approaches across six text classification tasks.