List your company

Long-form text generation on LongBench Write-en 1.0 (test)

91.8SqGPT-4o
51.96862.30972.6582.991Feb 4, 2025

Evaluation results

MethodLinks
GPT-4o2025.0291.8
GPT-4o mini2025.0290.3
LongWriter-Qwen + LongDPOBackbone=Qwen, Optimization=LongDPO2025.0288.6
Mistral-Large-Instruct2025.0288.3
LongWriter-Llama + LongDPOBackbone=Llama, Optimization=LongDPO2025.0288.2
Claude 3.5 Sonnet2025.0287.7
GPT-4 Turbo2025.0286.6
GLM-4-9B-chat2025.0285.5
LongWriter-LlamaBackbone=Llama2025.0282.2
Llama-3.1-70B-InstructParameters=70B2025.0280.3
Llama-3.1-8B-InstructParameters=8B2025.0270.6
Suri-I-ORPO2025.0253.5