AI Forever RAG - Multi-Hop Generation: leaderboard

The "Multi-Hop Generation" subtask of AI Forever RAG. A task from the AI Forever RAG benchmark, evaluating retrieval-augmented generation quality — answer faithfulness and relevance grounded in retrieved context, primarily for Russian-language models.

Metric: Average Generation Score. Source: huggingface.co. 2 models tracked.

Top models

#ModelScoreOverall rank
1gemma4-25B (kb graphRAG) (fe39ea)0.64
2Qwen3.8-27B (kb graphRAG) (aa045f)0.63

Interactive version: theaggregate.ai/benchmark?slug=ai-forever-rag-multi-hop-generation · How It Works · Data refreshed daily, snapshot 2026-10-20.