ViSP is a 1.2M-pair Vietnamese sentence paraphrase benchmark, generated by Gemini and human-validated, with baseline and LLM evaluations showing BARTpho-word large and Meta-Llama-3.1-70B lead their respective categories.
A lower Jaccard value indicates less overlap with the source, and hence greater paraphrase diversity
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CL 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
A Large-Scale Benchmark for Vietnamese Sentence Paraphrases
ViSP is a 1.2M-pair Vietnamese sentence paraphrase benchmark, generated by Gemini and human-validated, with baseline and LLM evaluations showing BARTpho-word large and Meta-Llama-3.1-70B lead their respective categories.