Hacker GitHub

← All repos

adamkarvonen/chess_llm_interpretability

View on GitHub ↗ ★ 221 Python

Visualizing the internal board state of a GPT trained on chess PGN strings, and performing interventions on its internal board state and representation of player Elo.

3 mentions on Hacker News by 3 people: 1 story, 2 comments · first 2024-01-06 · latest 2024-03-25

Mentions per year 3 mentions · busiest year 2024 (3)
20082026
AllStoriesComments NewestTopOldest
  1. The code for this is located here: https://github.com/adamkarvonen/chess_llm_interpretability
    comment by seraine · 2024-03-25 · in Manipulating Chess-GPT's World Model
  2. code here: https://github.com/adamkarvonen/chess_llm_interpretability
    comment by homarp · 2024-01-06 · in Chess-GPT's Internal World Model