Evaluate performance for 1-shot proof generation
Activity
abladek commentedon Oct 22, 2025
Code is working, llama3 failed every proof. Trying out different prompts to improve accuracy.
[-]N=1 Proof generation[/-][+]Proof generation[/+]abladek commentedon Oct 29, 2025
Working on multiple attempts
abladek commentedon Nov 4, 2025
Completed multiple attempts code with no improvement to success rate. 0% -> 0%. Runtime was ~2 hours with 5 attempts per theorem at 100 theorems.
abladek commentedon Nov 10, 2025
Can now amend proofs, tried with 100 proofs, 1 amend run, still 0% success rate, probably need to improve prompts.