Write and Improve 2024
A dataset of learner English essays based on the University of Cambridge's Write & Improve online writing platform, used to evaluate the generalization of NLI methods.
Timeline (last 90 days)
The study evaluated the token-probability scoring method using Llama 3.1 8B and 70B models on three datasets: TOEFL9, Write and Improve 2024, and BULATS-L1.
The 70B model achieved state-of-the-art (SOTA) performance on all datasets, surpassing prior prompt-based methods.
The Write and Improve 2024 dataset comes from the University of Cambridge's Write & Improve online writing platform and is relatively large in scale with diverse L1 backgrounds.
All Facts (3)
The study evaluated the token-probability scoring method using Llama 3.1 8B and 70B models on three datasets: TOEFL9, Write and Improve 2024, and BULATS-L1.
50%UnverifiedThe 70B model achieved state-of-the-art (SOTA) performance on all datasets, surpassing prior prompt-based methods.
50%UnverifiedThe Write and Improve 2024 dataset comes from the University of Cambridge's Write & Improve online writing platform and is relatively large in scale with diverse L1 backgrounds.
50%