XIFBench

XIFBench is a comprehensive constraint-based benchmark containing 558 multilingual instructions with varying constraints, used to evaluate the instruction-following abilities of large language models across different linguistic contexts.

Papers using XIFBench (3)

FAPO: Fully Autonomous Prompt Optimization of Multi-Step LLM Pipelines2026Ask, Don't Judge: Binary Questions for Interpretable LLM Evaluation and Self-Improvement2026XIFBench: Evaluating Large Language Models on Multilingual Instruction Following2025