Text Generation on AbGen

4.91Importance

ROMA

Updated 1mo ago

Evaluation Results

Method
ROMA 2026.02	4.91	4.64	4.62	4.93
GPT-4.1 2026.02	4.78	4.72	4.28	4.86
GPT-4o 2026.02	4.76	4.58	4.11	4.76
DeepSeek-V3.2-Exp 2026.02	4.74	4.75	4.17	4.79
deepseek-reasoner 2026.02	4.73	4.69	4.32	4.9
Qwen3-32B-AWQ 2026.02	4.72	4.43	4.19	4.73
Llama-4-Scout-17B-16E-Instruct 2026.02	4.69	4.1	4.01	4.38
Qwen3-235B-A22B-fp8-tput 2026.02	4.69	4.58	4.18	4.84
Qwen3-8B 2026.02	4.68	4.42	4.09	4.75
deepseek-chat 2026.02	4.68	4.54	4.1	4.77
Meta-Llama-3.1-70B-Instruct-FP8 2026.02	4.67	4.15	4.04	4.4
Mistral-Small-3.1-24B-Instruct-2503 2026.02	4.67	4.4	4.1	4.66
Qwen2.5-32B-Instruct 2026.02	4.67	4.17	4.07	4.61
o4-mini 2026.02	4.66	4.59	4.27	4.78
Gemini 2.5 Flash Preview-05-20 2026.02	4.65	4.64	4.15	4.66
DeepSeek-R1-0528-Qwen3-8B 2026.02	4.59	4.5	4.1	4.61
gemma-3-27b-it 2026.02	4.58	4.53	4.2	4.59
phi-4 2026.02	4.55	4.42	4.09	4.61
Llama-3.3-70B-Instruct-FP8-dynamic 2026.02	4.51	4.14	4.03	4.34
Llama-4-Maverick-17B-128E-Instruct-FP8 2026.02	4.5	4.42	4.03	4.56