D

Popular papers, already deconstructed.

A curated, public collection of research papers parsed with Deconstructed. Open any entry to explore its structure, equations, and AI explanations — no signup required.

  • Attention Is All You Need

    Vaswani et al.2017Transformers · NLP · Deep Learning

    The Transformer paper that replaced recurrence and convolutions with pure attention for sequence modeling. It became the architectural foundation for modern language models.

    arXiv
  • Language Models are Few-Shot Learners

    Brown et al.2020LLMs · Few-Shot Learning · Scaling

    The GPT-3 paper showing that scaling language models dramatically improves in-context and few-shot learning. It marked a major shift away from task-specific fine-tuning toward prompt-based use.

    arXiv
  • Learning to summarize from human feedback

    Stiennon et al.2020RLHF · Alignment · Summarization

    An early RLHF paper that improves summarization by training on human preferences instead of only matching reference summaries. It helped establish the practical value of preference-based optimization.

    arXiv
  • Denoising Diffusion Probabilistic Models

    Ho et al.2020Diffusion · Generative Models · Computer Vision

    A foundational diffusion paper showing high-quality image synthesis from iterative denoising. It helped kick off the modern diffusion wave in generative modeling.

    arXiv
  • Deep Residual Learning for Image Recognition

    He et al.2015Computer Vision · ResNets · Deep Learning

    Introduces ResNets and residual connections, making it practical to train much deeper neural networks. The paper became a cornerstone of modern computer vision and deep learning more broadly.

    arXiv
  • LoRA: Low-Rank Adaptation of Large Language Models

    Hu et al.2021LLMs · Fine-Tuning · Parameter Efficiency

    Introduces LoRA, a parameter-efficient fine-tuning method that freezes base weights and learns small low-rank updates. It became one of the standard techniques for adapting large models cheaply.

    arXiv
  • BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding

    Devlin et al.2018NLP · Pre-training · Transformers

    Introduces BERT, a bidirectional Transformer pre-trained with masked language modeling. It reset the standard for transfer learning in NLP and powered a huge wave of downstream fine-tuning work.

    arXiv
  • Learning Transferable Visual Models From Natural Language Supervision

    Radford et al.2021CLIP · Vision-Language · Zero-Shot

    The CLIP paper showing that image-text pretraining produces highly transferable visual representations. It demonstrated strong zero-shot performance across a wide range of vision tasks.

    arXiv
  • Semi-Supervised Classification with Graph Convolutional Networks

    Kipf and Welling2016Graph Neural Networks · Semi-Supervised Learning · Representation Learning

    A landmark GCN paper that brought efficient graph convolutions into mainstream machine learning. It became one of the canonical starting points for graph representation learning.

    arXiv