Multi-IF

The 'Multi-IF' dataset/benchmark is used to evaluate the performance of models in out-of-domain tasks within the context of agentic reinforcement learning.

Papers using Multi-IF (3)

Planner-R1: Reward Shaping Enables Efficient Agentic RL with Smaller LLMs2025Reverse Preference Optimization for Complex Instruction Following2025Multi-IF: Benchmarking LLMs on Multi-Turn and Multilingual Instructions Following2024 · 5 cites