100-point leaderboard
#4
by
maximazzik
- opened
How reasonable is it that the top solutions scoring 100 points simply retrieve correct answers from... a database of correct answers? Sure, the provided LLM toolkit includes various tools, but the answers are still being pulled directly from "supabase_docs.csv", turning an agent-based competition into a RAG task.
This is not fair indeed. I am working on tools that currently do the tasks without prior knowledge of correct answers to tackle the challenge properly!