圣达菲

conda activate ds2_py311

scp -r yuanyi@192.168.1.4:/home/yuanyi/DS2/LLM_scoring/scoring.sh /mnt/c/users/zhangy85/Desktop/

export AZURE_OPENAI_KEY="8EHOOtsZ2OugjLJCbB5mMjigc24DMl8TkExITvoQJQVEY2T1hsA4JQQJ99BGAC4f1cMXJ3w3AAABACOG1Myz"

export AZURE_OPENAI_ENDPOINT="https://dsproblema.openai.azure.com/"

先降噪

去整, choosen和reject 丢到里面, 算一下noisy和transition matrix

figure3的图 correlated score

先只对choosen弄图, 再弄reject的图, 最后结合两个一起算

应该是对角线全是1的matrix

先看看数据比例, 看看要不要 merge 数据. 看看merge成几类比较合适, 都试试

conda activate ds2

conda activate ds2_py311

scp -r yuanyi@192.168.1.4:/home/yuanyi/DS2/LLM_scoring/scoring.sh /mnt/c/users/zhangy85/Desktop/

export AZURE_OPENAI_KEY="8EHOOtsZ2OugjLJCbB5mMjigc24DMl8TkExITvoQJQVEY2T1hsA4JQQJ99BGAC4f1cMXJ3w3AAABACOG1Myz"

export AZURE_OPENAI_ENDPOINT="https://dsproblema.openai.azure.com/"

先降噪

去整, choosen和reject 丢到里面, 算一下noisy和transition matrix

figure3的图 correlated score

先只对choosen弄图, 再弄reject的图, 最后结合两个一起算

应该是对角线全是1的matrix

先看看数据比例, 看看要不要 merge 数据. 看看merge成几类比较合适, 都试试

conda activate ds2