Research Library

Library

Papers, patents & whitepapers, annotated

Whitepaper2023

GPT-4 Technical Report

OpenAI, Josh Achiam, Steven Adler et al. · OpenAI (Technical Report) · 2023

The moment frontier reports stopped being reproducible science and became capability-and-safety disclosures. Worth reading alongside an open report like DeepSeek's or OLMo's to see exactly which questions each one refuses to answer.

Whitepaper2022

Constitutional AI: Harmlessness from AI Feedback

Yuntao Bai, Saurav Kadavath, Sandipan Kundu et al. · Anthropic (Technical Report) · 2022

Notable for making the values explicit and auditable rather than implicit in whichever labels the annotators happened to produce. The self-critique loop is also the ancestor of the LLM-as-judge pipelines now used far beyond safety work.

Whitepaper2022

BLOOM: A 176B-Parameter Open-Access Multilingual Language Model

BigScience Workshop, :, Teven Le Scao et al. · BigScience (Technical Report) · 2022

The first serious demonstration that a frontier-scale model could be built outside a private lab. Its language coverage remains unusual — most open models are still overwhelmingly English.

Paper2022

FlashAttention: Fast and Memory-Efficient Exact Attention with IO-Awareness

Tri Dao, Daniel Y. Fu, Stefano Ermon et al. · NeurIPS · 2022

The rare optimisation with no accuracy trade-off to weigh, which is why it was absorbed everywhere within a year. It is also the clearest lesson that on modern hardware the bottleneck is usually memory movement, not compute.

Preprint2022

Training Compute-Optimal Large Language Models

Jordan Hoffmann, Sebastian Borgeaud, Arthur Mensch et al. · arXiv · 2022

The result that redirected the field from 'bigger' to 'better fed', and the reason the models that followed got smaller and cheaper to serve while getting better. If you read one scaling paper, read this one.

Paper2022

Training language models to follow instructions with human feedback

Long Ouyang, Jeff Wu, Xu Jiang et al. · NeurIPS · 2022

The single most consequential result on this shelf for how AI products actually feel. It established that instruction-following is a separate axis from scale — and made the case that a smaller aligned model can beat a much larger unaligned one.

Paper2022

Chain-of-Thought Prompting Elicits Reasoning in Large Language Models

Wei, Wang, Schuurmans et al. · NeurIPS · 2022

Why "let's think step by step" works. The practical root of every reasoning-heavy prompt and the reasoning-model wave that followed.

Patent2021

Processing sequences using convolutional neural networks

DeepMind Technologies Ltd · USPTO — US11080591B2 · 2021

The convolutional answer to sequence modelling, filed as the transformer was arriving. A good illustration that the architecture that wins is not the only one that worked.

Paper2021

LoRA: Low-Rank Adaptation of Large Language Models

Edward J. Hu, Yelong Shen, Phillip Wallis et al. · ICLR · 2021

The reason fine-tuning is something an individual can do on one GPU rather than a datacentre operation. Nearly every 'custom model' offering is this underneath.

Paper2021

Learning Transferable Visual Models From Natural Language Supervision

Alec Radford, Jong Wook Kim, Chris Hallacy et al. · ICML · 2021

Where zero-shot became a practical expectation rather than a research curiosity. The shared image-text space it introduced is still the substrate for multimodal search and for how most image generators are conditioned.

Paper2021

Switch Transformers: Scaling to Trillion Parameter Models with Simple and Efficient Sparsity

William Fedus, Barret Zoph, Noam Shazeer · J. Mach. Learn. Res. · 2021

The architecture behind most frontier models today, where a very large parameter count coexists with a modest per-token cost. Read it before the Kimi and DeepSeek reports here — it is the idea they are both building on.

Preprint2020

Language Models are Few-Shot Learners

Brown, Mann, Ryder et al. · arXiv · 2020

The GPT-3 paper — where "just give it a few examples in the prompt" became a real strategy. It reframed prompting as programming.

Paper2020

Retrieval-Augmented Generation for Knowledge-Intensive NLP Tasks

Lewis, Perez, Piktus et al. · NeurIPS · 2020

The origin of "RAG". Explains why bolting a retriever onto a model beats fine-tuning for fast-changing knowledge — the pattern behind most production AI search today.

Patent2020

Multi-task learning using knowledge distillation

Google LLC · USPTO — US10635977B2 · 2020

Distillation as claimed intellectual property, and the technique behind nearly every small model that punches above its size — including several on this shelf. Read it next to the on-policy distillation work here to see how far the idea has since travelled.

Paper2020

Dense Passage Retrieval for Open-Domain Question Answering

Vladimir Karpukhin, Barlas Oğuz, Sewon Min et al. · EMNLP · 2020

The retrieval half of RAG, and the reason vector search became the default. Read it with the RAG paper here — this is the component that made that architecture work.

Paper2019

Exploring the Limits of Transfer Learning with a Unified Text-to-Text Transformer

Colin Raffel, Noam Shazeer, Adam Roberts et al. · J. Mach. Learn. Res. · 2019

Valuable as a controlled study rather than a model release: it is one of the few places where pretraining decisions are compared like-for-like instead of confounded. The text-to-text framing is now simply how everyone uses language models.

Patent2019

Attention-based sequence transduction neural networks

Google LLC · USPTO — US10452978B2 · 2019

The Transformer, as a granted patent. Useful for seeing how a landmark architecture is framed in IP terms — and a reminder that the research and the patent often diverge.

Paper2018

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

Jacob Devlin, Ming-Wei Chang, Kenton Lee et al. · NAACL-HLT · 2018

The paper that made 'pretrain once, fine-tune cheaply' the default. Encoder models like this still underpin most retrieval and classification stacks, including the reranking step in a typical RAG pipeline.

Paper2017

Attention Is All You Need

Vaswani, Shazeer, Parmar et al. · NeurIPS · 2017

The paper the entire modern LLM era is built on. If you read one source here, read this one — every model you use is a descendant of this architecture.

Patent2016

System and method for addressing overfitting in a neural network

Google LLC · USPTO — US9406017B2 · 2016

Dropout, patented. Worth a look because it is one of the clearest cases of a technique that spread through the field as a free idea while also existing as an enforceable claim.

Showing 4160 of 66 sources · newest first