Question Answering on HolisQA Computer Science
31.6F1HugRAG
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| HugRAGBackbone LLM=gpt-5-nano2026.02 | 31.6 | 60.94 | 58.34 | |
| CausalRAGBackbone LLM=gpt-5-nano2026.02 | 30.98 | 54.1 | 57.54 | |
| LeanRAGBackbone LLM=gpt-5-nano2026.02 | 30.51 | 57.61 | 55.45 | |
| StandardRAGBackbone LLM=gpt-5-nano2026.02 | 28.87 | 49.44 | 57.1 | |
| GraphRAG LocalBackbone LLM=gpt-5-nano2026.02 | 25.1 | 39.9 | 53.3 | |
| BM25Backbone LLM=gpt-5-nano2026.02 | 24 | 39.12 | 52.4 | |
| GraphRAG GlobalBackbone LLM=gpt-5-nano2026.02 | 23.75 | 37.65 | 53.17 | |
| LightRAGBackbone LLM=gpt-5-nano2026.02 | 22.59 | 41.86 | 51.62 | |
| NaiveGenerationBackbone LLM=gpt-5-nano2026.02 | 18.93 | 0 | 48.79 | |
| HippoRAG2Backbone LLM=gpt-5-nano2026.02 | 16.94 | 21.05 | 47.29 |