README.md

metadata

license: cc-by-nc-4.0
language:
  - en
tags:
  - instruction following
  - multi-turn
  - LLM
  - rubric-based

Dataset Summary

We introduce AdvancedIF, a new benchmark featuring over 1,600 prompts and expert-curated rubric designed to assess LLMs' proficiency in

• Complex instruction following: each prompt has 6+ instructions with combination of one, format, style, structure, length, negative constraints, spelling, and inter-conditional instructions;
• Multi-turn instruction following: the ability to follow instruction carried from previous;
• System prompt steerability: The ability to follow instructions in the system prompt.

See paper for full details:

AdvancedIF: Rubric-Based Benchmarking and Reinforcement Learning for Advancing LLM Instruction Following

Evaluation Script

https://github.com/facebookresearch/AdvancedIF

Data Splits

• test: 1,645 examples