seraine

3 ms·

1. 1.▲
   LLM inference is nearly deterministic. We use this to audit providers(adamkarvonen.github.io)

   1 points by seraine 11mo ago|0 comments
2. 2.▲
   by seraine 3y ago

   The Stockfish program can be set to play at strength level 0-20. Estimates of the levels' Elo is provided here: https://github.com/official-stockfish/Stockfish/commit/a08b8...
3. 3.▲
   by seraine 3y ago

   I updated the article to also mention this in the second blog post. All games were against Stockfish level 0, with nodes searched limited to 100,000 rather than a time based search limit to limit variability due to different processors.
4. 4.▲
   Manipulating Chess-GPT's World Model(adamkarvonen.github.io)

   119 points by seraine 3y ago|36 comments
5. 5.▲
   by seraine 3y ago

   The code for this is located here: https://github.com/adamkarvonen/chess_llm_interpretability
6. 6.▲
   by seraine 3y ago

   What did you use to play against it?
7. 7.▲
   by seraine 3y ago

   Correct. An example can be seen here: https://github.com/adamkarvonen/chess_gpt_eval/blob/master/g...
8. 8.▲
   GPT-3.5-instruct is a ~1800 ELO chess player. 150 games of GPT-3.5 vs. stockfish(twitter.com)

   3 points by seraine 3y ago|2 comments