- [P] two copies of gpt-3.5 (one playing as the oracle, and another as the guesser) performs poorly on the game of 20 Questions (68/1823). https://evanthebouncy.medium.com/llm-self-play-on-20-questions-dee7a8c63377 20 comments machinelearning
Linked pages
Would you like to stay up to date with Computer science? Checkout Computer science
Weekly.
Related searches:
Search whole site: site:evanthebouncy.medium.com
Search title: LLM self-play on 20 Questions. gpt-3.5-turbo has a score of 68/1823… | by Evan Pu | Mar, 2023 | Medium
See how to search.