R0027/2026-03-26/Q001/SRC02
Mondshine et al. — Prompt translation strategies across 35 languages
Source
| Field |
Value |
| Title |
Beyond English: The Impact of Prompt Translation Strategies across Languages and Tasks in Multilingual LLMs |
| Publisher |
arXiv / LoResMT 2025 |
| Author(s) |
Itai Mondshine, Tzuf Paz-Argaman, Reut Tsarfaty |
| Date |
2025-02 |
| URL |
https://arxiv.org/html/2502.09331v1 |
| Type |
Research paper |
Summary
| Dimension |
Rating |
| Reliability |
High |
| Relevance |
High |
| Bias: Missing data |
Low risk |
| Bias: Measurement |
Low risk |
| Bias: Selective reporting |
Low risk |
| Bias: Randomization |
N/A — not an RCT |
| Bias: Protocol deviation |
N/A — not an RCT |
| Bias: COI/Funding |
Low risk |
Rationale
| Dimension |
Rationale |
| Reliability |
Rigorous methodology: 35 languages, 4 tasks, 24 prompt configurations per language/task, multiple models. Bar-Ilan University affiliation. |
| Relevance |
Directly answers whether prompt language affects performance and quantifies the effect across high/medium/low resource languages. |
| Bias flags |
Low risk across domains. Transparent methodology with Association Rule Learning analysis. |
| Evidence ID |
Summary |
| SRC02-E01 |
Selective pre-translation consistently outperforms full translation and native inference |
| SRC02-E02 |
Low-resource languages show 200%+ improvement with selective pre-translation |