Manipulating Chess-GPT’s World Model | Adam Karvonen

Manipulating Chess-GPT’s World Model

Manipulating Chess-GPT’s World Model Note: This work has since been turned into a paper accepted to the Conference on Language Modeling , but the average reader will probably prefer the blog post. In my previous post I introduced Chess-GPT, a language model I trained to predict the next character in a game of chess given a PGN string (1.e4 e5 2.Nf3 …). Through the process of training to output the next character, it learns to compute the state of the chess board and to estimate the skill level of the players in the game given an arbitrary PGN string as input. I demonstrated this using linear p

saved by

• Akshith Garapati

related reading

• Chess-GPT’s Internal World Model | Adam Karvonenadamkarvonen.github.io
• Evaluating Sparse Autoencoders with Board Games | Adam Karvonenadamkarvonen.github.io
• [2210.13382] Emergent World Representations: Exploring a Sequence Model Trained on a Synthetic Taskarxiv.org
• Actually, Othello-GPT Has A Linear Emergent World Representation - Neel Nandaneelnanda.io
• Playing chess with large language modelsnicholas.carlini.com
• Large Language Model: world models or surface statistics?thegradient.pub
• Actually, Othello-GPT Has A Linear Emergent World Representation — AI Alignment Forumalignmentforum.org
• LLM Chess: Benchmarking Reasoning and Instruction-Following in LLMs through Chessarxiv.org
• gpt-4.pdfcdn.openai.com
• A Very Unlikely Chess Game | Slate Star Codexslatestarcodex.com
• Chapter 1: Transformer Interpretability - ARENAlearn.arena.education
• OthelloGPT learned a bag of heuristics — LessWronglesswrong.com