Architecture foundations
AlexNet → ResNet → Transformer → Vision Transformer traces convolution, residual connections, and attention as model backbones.
AI research paper reading path
20 papers that explain how modern AI evolved
For a dependable map of modern AI, start with AlexNet, ResNet, and the Transformer for architecture; continue through BERT, GPT-3, RAG, and InstructGPT for language models; then add diffusion, multimodal systems, and agents. This is not a mechanical citation ranking. Each paper changed an architecture, a training recipe, or the products researchers could build.
Choose a path for your goal
Large language models
BERT → Scaling Laws → GPT-3 → RAG → InstructGPT → Chinchilla covers pre-training, scale, retrieval, and alignment.
Generation and agents
VAE → GAN → DDPM → Latent Diffusion, then Chain-of-Thought and ReAct, connects generation, reasoning, and action.
The 20 papers and what each one changed
Every entry links to the primary paper and, when available, a plain-English PaperBridge visual explainer.
- 01
ImageNet Classification with Deep Convolutional Neural Networks
It joined GPUs, large datasets, and deep convolutional networks into a scalable vision recipe.
- 02
Auto-Encoding Variational Bayes
It made probabilistic latent-variable models trainable with ordinary backpropagation.
- 03
Generative Adversarial Networks
It reframed generation as a game between a generator and a discriminator.
- 04
Deep Residual Learning for Image Recognition
Residual connections made substantially deeper networks practical to optimize.
- 05
Attention Is All You Need
The Transformer replaced recurrence with attention and became the shared backbone of modern foundation models.
- 06
BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding
It established bidirectional pre-training followed by task fine-tuning as a general NLP recipe.
- 07
Scaling Laws for Neural Language Models
It showed that model loss follows predictable power laws across scale, data, and compute.
- 08
Language Models are Few-Shot Learners
It demonstrated broad few-shot task learning through prompting at unprecedented scale.
- 09
Retrieval-Augmented Generation for Knowledge-Intensive NLP Tasks
It combined parametric memory with retrievable external knowledge, shaping modern knowledge assistants.
- 10
Denoising Diffusion Probabilistic Models
It turned iterative denoising into a stable route to high-quality generation.
- 11
An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale
It showed that a pure Transformer over image patches could outperform convolutional networks at scale.
- 12
Learning Transferable Visual Models From Natural Language Supervision
It aligned images and text at web scale, enabling strong zero-shot visual transfer.
- 13
LoRA: Low-Rank Adaptation of Large Language Models
It adapted large models through small low-rank updates instead of retraining every weight.
- 14
High-Resolution Image Synthesis with Latent Diffusion Models
It moved diffusion into a compressed latent space, making high-resolution text-to-image generation practical.
- 15
Chain-of-Thought Prompting Elicits Reasoning in Large Language Models
It showed that examples with intermediate steps can unlock much stronger reasoning in large models.
- 16
Training language models to follow instructions with human feedback
It established a practical recipe for aligning language models with human instructions and preferences.
- 17
Training Compute-Optimal Large Language Models
It showed that many large models were under-trained and reset the compute-optimal balance of parameters and data.
- 18
ReAct: Synergizing Reasoning and Acting in Language Models
It interleaved reasoning with tool actions, providing a direct blueprint for modern agent loops.
- 19
Segment Anything
It turned segmentation into a promptable, transferable foundation capability backed by a large dataset.
- 20
Visual Instruction Tuning
It connected a vision encoder and a language model through visual instruction tuning to build a conversational assistant.
FAQ
Essential AI paper FAQ
Which AI paper should a beginner read first?
Start with Attention Is All You Need for the Transformer, then choose BERT, RAG, or diffusion according to your goal. Use a first pass focused on the problem, method, evidence, and limitations.
Do I need to read all 20 papers in full?
No. Product readers can begin with the abstract, method figure, main result, and limitations. Engineers can continue into equations, training details, baselines, and ablations.
Why are some famous model papers missing?
The list favors turning points that explain later research rather than ranking brands or recent popularity. It is a foundation map, not a complete history.