Wow, was surprised to see that grading hacks did not teach hacking. Why was Qwen3-14B the model chosen for this project, do you think you would get a different conclusion depending on what LLM model was used?
Very insightful article, great breakdown on the value leakage problem!
Wow, was surprised to see that grading hacks did not teach hacking. Why was Qwen3-14B the model chosen for this project, do you think you would get a different conclusion depending on what LLM model was used?