BERTScore

anchor

Text similarity measured by matching contextual embeddings between candidate and reference rather than by overlapping tokens.

Findings

• IC-027LMM performance in deciphering oracle bone inscriptions is comparable to untrained humans for common characters but declines for rarer and structurally complex characters [eval]
• IC-304Instruction-tuned LMs become more vulnerable to prompt-injected data extraction as model size increases from 7B to 70B [eval]
• IC-812InstructGPT (text-davinci-003) reduces content diversity in co-written essays while GPT-3 (davinci) does not, and the effect is attributable to the model's own less diverse text contributions [eval]