Long-form Retrieval-Augmented Generation on LongEval Task 4 (primary)
16.7ROUGE-Ltopic-shift-current
Evaluation Results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| topic-shift-currentChunking=topic-shift, Reranker=current2026.07 | 16.7 | 0.156 | 92.9 | 29.2 | 2.694 | 0.215 | |
| semantic-minilmChunking=semantic, Reranker=MiniLM2026.07 | 16 | 0.157 | 92.2 | 29.1 | 2.696 | 0.349 | |
| rule-minilmChunking=rule-based, Reranker=MiniLM2026.07 | 15.7 | 0.168 | 97.5 | 36.7 | 2.929 | 0.305 | |
| topic-shift-minilmChunking=topic-shift, Reranker=MiniLM2026.07 | 15.7 | 0.163 | 94 | 28.8 | 2.671 | 0.373 | |
| semantic-currentChunking=semantic, Reranker=current2026.07 | 15.6 | 0.15 | 93.3 | 28.2 | 2.651 | 0.262 | |
| single-query-bm25Retrieval=BM252026.07 | 15.6 | 0.152 | 94 | 30.8 | 2.744 | 0.12 | |
| rrf-no-rerankRank Aggregation=RRF, Reranking=none2026.07 | 15.5 | 0.151 | 92.9 | 33.6 | 2.816 | 0.087 | |
| caes-rag-rrfFramework=CAES-RAG, Rank Aggregation=RRF2026.07 | 15.5 | 0.144 | 94.2 | 33.8 | 2.83 | 0.059 | |
| default2026.07 | 14.7 | 0.146 | 93.1 | 33.4 | 2.845 | 0.025 | |
| official-naive-baseline2026.07 | 8.2 | -0.118 | 20 | 12.4 | 1.828 | -0.758 |