Safety
Multi-SFT: Permissively-Licensed Multilingual Instruction-Following
arXiv:2610.22086v1 Announce Type: cross Abstract: Supervised fine-tuning datasets remain overwhelmingly English-centric and commercially restricted, as their content is frequently generated by proprie
arXiv:2610.22086v1 Announce Type: cross Abstract: Supervised fine-tuning datasets remain overwhelmingly English-centric and commercially restricted, as their content is frequently generated by proprietary models whose terms of service forbid training competing systems. We introduce Multi-SFT, a commercially permissive collection of 502,818 instruction-response pairs across five European languages, built via translation and regeneration pipelines for tasks where translation is insufficient. Human evaluation yields inter-annotator agreement of at least 0.87 across all datasets and languages. Fine-tuning nine models from three different families (EuroLLM, OLMo 2, Qwen3; 1B-32B) and evaluating them on 13 benchmarks shows that Multi-SFT alone is competitive with larger augmented mixes and official instruction-tuned releases, with decisive advantages on safety tasks.
Related
- Routing-Aligned Fine-Tuning for Multilingual Downstream Tasks in Mixture-of-Experts Models
- Beyond Bilingual Transfer: Multilingual Code-Switching in Instruction Tuning
- Step-Level On-Policy Distillation: Interpolating Between On-Policy Distillation and Supervised Fine-Tuning
- Goal-Conditioned Supervised Learning for LLM Fine-Tuning
Source: arXiv cs.AI | 2026-10-16