Summarization on SummaryBench Roman (FR, WC (s))
100Fidelity Rate (FR)Direct
Evaluation results
| Method | Links | ||
|---|---|---|---|
| DirectLLM Backend=GPT-5.4 mini2026.05 | 100 | — | |
| Magentic-OneLLM Backend=GPT-5.4 mini2026.05 | 100 | — | |
| MegaAgentLLM Backend=GPT-4.1 mini2026.05 | 70 | 579 | |
| APWALLM Backend=GPT-5.4 mini2026.05 | 0 | 329 |