Is the high drop in GSM8K usual?

#1
by dball - opened
Owner

Overall comparison shows that most metrics improve through DPO:

dball/zephyr-7b-dpo-qlora: (+)Average: 61.27; -GSM8K 33.97; (+)Win 78.61; +TQA: 44.03; -+MMLU 62.28;+ HSwag 84.92; +ARC 63.82
dball/zephyr-7b-sft-qlora (-)Average: 59.8; -GSM8k 34.12; (-)Win 78.22; (+)TQA 42.32; -MMLU 61.9; (-)HSwag 82.49; (-)ARC 59.73
mistralai/Mistral-7B-v0.1: Average: 60.97; GSM8K 37.83; Win 78.37; TQA: 42.15; MMLU 64.16; HSwag 83.31; ARC 59.98

Sign up or log in to comment