📰 Lilian Weng's Blog

Special thanks to John Schulman for a lot of super valuable feedback and direct edits on this post. Test time compute ( Graves et al. 2016 , Ling, et al. 2017 ,

Reward hacking occurs when a reinforcement learning (RL) agent exploits flaws or ambiguities in the reward function to achieve high rewards, without genuinely l

Hallucination in large language models usually refers to the model generating unfaithful, fabricated, inconsistent, or nonsensical content. As a term, hallucina

Diffusion models have demonstrated strong results on image synthesis in past years. Now the research community has started working on a harder task—using

[Special thank you to Ian Kivlichan for many useful pointers (E.g. the 100+ year old Nature paper “Vox populi”) and nice feedback. 🙏 ] High-quality

The use of large language models in the real world has strongly accelerated by the launch of ChatGPT. We (including my team at OpenAI, shoutout to them) have in

Building agents with LLM (large language model) as its core controller is a cool concept. Several proof-of-concepts demos, such as AutoGPT , GPT-Engineer and Ba

Prompt Engineering , also known as In-Context Prompting , refers to methods for how to communicate with LLM to steer its behavior for desired outcomes without u

Many new Transformer architecture improvements have been proposed since my last post on “The Transformer Family” about three years ago. Here I did a

[Updated on 2023-01-24: add a small section on Distillation .] Large transformer models are mainstream nowadays, creating SoTA results for a variety of tasks. T

Neural networks are well known to be over-parameterized and can often easily fit data with near-zero training loss with decent generalization performance on tes

Processing images to generate text, such as image captioning and visual question-answering, has been studied for years. Traditionally such systems rely on an ob

Here comes the Part 3 on learning with not enough data (Previous: Part 1 and Part 2 ). Let’s consider two approaches for generating synthetic data for training.

This is part 2 of what to do when facing a limited amount of labeled data for supervised learning tasks. This time we will get some amount of human labeling wor

When facing a limited amount of labeled data for supervised learning tasks, four appr

[Updated on 2022-03-13: add expert choice routing .] [U

[Updated on 2021-09-19: Highly recommend this blog post on score-based generative modeling

The goal of contrastive representation learning is to learn such an embedding space in which similar sample pairs stay close to each other while dissimilar ones

Large pretrained language models are trained over a sizable collection of online data. They unavoidably acquire certain toxic behavior

[Updated on 2021-02-01: Updated to version 2.0 with several wor

[Updated on 2020-11-12: add an example on closed-book factual QA using OpenAI API (beta).

Although most popular and successful model architectures are designed by hum

[Updated on 2020-06-17: Add “exploration via disagreement” in the “Forward Dynamics” section . Exploitation versus ex

[Updated on 2023-01-27 : After almost three years, I did a big refactoring update of this post to incorporate a bunch of new Transformer models since 2020. The