Pre-finetuning without key phrases does not improve scoring performance and slightly lowers it relative to the BERT baseline¶

Claim · All claims
Evidence · 1 study · causal r2 · q2
Currency · newest study 2025 · not yet checked against its sources · not reviewed by a person

Subclaims¶

q2 i? Simply pre-finetuning on other prompts without key-phrase inputs slightly lowered model performance compared with the BERT baseline. → Funayama 2025

Evidence¶

Funayama 2025¶

Funayama, H., Matsubayashi, Y., Asazuma, Y., Mizumoto, T., & Inui, K. (2025). Cross-prompt Pre-finetuning of Language Models for Short Answer Scoring. Int J Artif Intell Educ. https://doi.org/10.1007/s40593-025-00474-w

q2 · i? · causal · r2

In the same five-setting BERT experiment on the RIKEN dataset, the 2-phase setting without key phrases underperformed the baseline, and using key phrases without pre-finetuning also did not improve performance. The authors attribute this to key phrases remaining "unchanged throughout the training phase, thus failing to provide any dynamic cues."

"2-phase w/o key phrase slightly lowered the model performance compared with the BERT baseline. This result indicates that simply pre-finetuning on other prompts cannot effectively improve model performance."

Discussion¶

Related Claims¶

• Two-phase cross-prompt pre-finetuning with key phrases significantly improves short answer scoring accuracy when in-prompt training data is limited — related
• The pre-finetuned model learns a scoring principle: predicted scores correlate strongly and negatively with edit distance between answers and key phrases — related
• Pre-finetuning performance increases with the number of distinct prompts used, holding total pre-finetuning answers fixed — related
• Cross-prompt pre-finetuning shows no significant benefit when sufficient (200-answer) in-prompt training data is available — related