8

monkey_project

Kaggle Usernames

concl Zandy Zhao Marissa Li ABigNerd werus23

Docs

(cv-lb spreadsheet)[https://docs.google.com/spreadsheets/d/15m7-UEgd3Te7h3DRnemhhly0h5DYlFFvWIKELjPfsU0/edit?usp=sharing]

Notebooks

(Use embeddings + LLMs to guess misconceptions)[https://www.kaggle.com/werus23/eedi-few-shot-w-qwen2-5-32b-instruct-awq-vllm]

(play around with postprocessing-ranking misconceptions)[https://www.kaggle.com/werus23/eedi-few-shot-inspection]

Tasks

check misconceptions represented equally in training data (they are not)

improve few shot prompting with embeddings (did it)

finetune embedding model

finetune LLM misconception generator?

check for duplicate misconceptions