Dynamic scheduling on DynaSchedBench
1.01Mean CmaxQwen3-8B
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| Qwen3-8BType=LLM2026.05 | 1.01 | 0.2173 | |
| Best heuristicType=Heuristic2026.05 | 1.11 | 0.154 | |
| Qwen3-14BType=LLM2026.05 | 1.34 | 0.1955 | |
| Qwen3-4BType=LLM2026.05 | 1.45 | 0.2163 | |
| Kimi-K2Type=LLM2026.05 | 1.46 | 0.2383 | |
| Claude-4.5 HaikuType=LLM2026.05 | 1.52 | 0.2342 | |
| Qwen3-32BType=LLM2026.05 | 1.65 | 0.2072 | |
| DeepSeek-V3.2Type=LLM2026.05 | 1.68 | 0.2943 | |
| Gemini-2.5 FlashLiteType=LLM2026.05 | 1.76 | 0.2921 | |
| Grok-4 FastType=LLM2026.05 | 1.88 | 0.2316 | |
| GPT-5 MiniType=LLM2026.05 | 1.93 | 0.2469 | |
| Median heuristicType=Heuristic2026.05 | 1.94 | 0.2479 | |
| Worst heuristicType=Heuristic2026.05 | 51.52 | 3.8762 |