id,setting,family,K,predicted,base,ta,prism
qwen1.5b_ja_med_cpt,"Qwen2.5-1.5B ja + med, constructed",Qwen2.5,2,1,56.1,35.8,55.9
qwen1.5b_math_coder,Qwen2.5-1.5B Math + Coder,Qwen2.5,2,0,56.3,37.1,55.8
qwen7b_math_coder,Qwen2.5-7B Math + Coder,Qwen2.5,2,0,68.0,49.5,67.8
qwen7b_r1_coder,Qwen2.5-7B R1-Distill + Coder,Qwen2.5,2,1,68.1,49.7,67.7
qwen7b_prm_coder,Qwen2.5-7B Math-PRM + Coder,Qwen2.5,2,1,68.1,53.7,67.6
qwen7b_math_finance,Qwen2.5-7B Math + Finance,Qwen2.5,2,1,68.1,37.0,
qwen7b_coder_coderinst,Qwen2.5-7B Coder + Coder-Instruct,Qwen2.5,2,1,68.1,39.3,
deepseek7b_math_chat,DeepSeek-7B math + chat,DeepSeek,2,0,46.8,48.1,46.6
qwen7b_selfsft_10x_lr,Qwen2.5-7B self-SFT at 10x learning rate,Qwen2.5,2,1,68.1,68.8,68.2
llama8b_r1_tulu,Llama-3.1-8B R1-Distill + Tulu-3,Llama-3.1,2,1,59.3,62.6,59.6
qwen7b_aloe_math,Qwen2.5-7B Aloe medical + Math,Qwen2.5,2,1,68.1,68.0,68.1
qwen7b_huatuo_r1,Qwen2.5-7B HuatuoGPT + R1-Distill,Qwen2.5,2,1,68.1,68.9,68.1
qwen7b_math_coder_instruct,Qwen2.5-7B Math + Coder + Instruct,Qwen2.5,3,0,68.0,68.6,68.1
mistral7b_metamath_code,Mistral-7B MetaMath + Code,Mistral,2,0,57.4,59.9,
qwen7b_selfsft_k2,"Qwen2.5-7B self-SFT, K = 2",Qwen2.5,2,0,68.1,68.4,68.2
mistral7b_saul_bio,Mistral-7B SaulLM legal + BioMistral,Mistral,2,1,57.3,57.8,57.4
qwen7b_math_eva,Qwen2.5-7B Math + EVA roleplay,Qwen2.5,2,1,68.1,68.1,68.1
llama8b_swallow_guard,Llama-3.1-8B Swallow ja + Guard,Llama-3.1,2,1,60.2,60.0,59.7
llama8b_openmath_hermes_guard,Llama-3.1-8B OpenMath2 + Hermes-3 + Guard,Llama-3.1,3,0,60.2,60.0,59.7
qwen7b_offtheshelf_k6,Qwen2.5-7B off-the-shelf x6,Qwen2.5,6,0,68.1,68.0,68.2
qwen7b_selfsft_k8,"Qwen2.5-7B self-SFT, K = 8",Qwen2.5,8,0,68.1,68.0,68.0
llama8b_tulu_dpo,Llama-3.1-8B Tulu-3 + Tulu-3-DPO,Llama-3.1,2,1,59.3,64.5,