API Usage on APIBench HF (test)
83.2AccuracyExpeL + Janus
Evaluation Results
| Method | Links | |
|---|---|---|
| ExpeL + JanusBackbone LLM=DeepSeek-V4-Flash2026.06 | 83.2 | |
| DC-RS + JanusBackbone LLM=Qwen3-8B2026.06 | 82.8 | |
| ExpeLBackbone LLM=DeepSeek-V4-Flash2026.06 | 82 | |
| DC-RSBackbone LLM=Qwen3-8B2026.06 | 80.8 | |
| DC-RS + JanusBackbone LLM=DeepSeek-V4-Flash2026.06 | 76.8 | |
| DC-RSBackbone LLM=DeepSeek-V4-Flash2026.06 | 74.8 | |
| ExpRAGBackbone LLM=Qwen3-8B2026.06 | 72 | |
| ExpRAGBackbone LLM=DeepSeek-V4-Flash2026.06 | 71.6 | |
| ExpeL + JanusBackbone LLM=Qwen3-8B2026.06 | 70.4 | |
| Memory-freeBackbone LLM=DeepSeek-V4-Flash2026.06 | 66.8 | |
| ExpeLBackbone LLM=Qwen3-8B2026.06 | 66.4 | |
| Memory-freeBackbone LLM=Qwen3-8B2026.06 | 65.2 |