Training Speedup on Campaign 1
27.8Speedup (%)Aletheia
Evaluation results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| AletheiaModel=Qwen2.5-14B2026.04 | 27.8 | — | 140.9 | 1.385 | 0.0001 | |
| AletheiaModel=Mistral-7B2026.04 | 27.7 | — | 96.2 | 1.383 | 0.0002 | |
| AletheiaModel=Qwen2.5-1.5B2026.04 | 27.1 | — | 82.4 | 1.371 | 0 | |
| AletheiaModel=Llama-3.1-8B2026.04 | 26.7 | — | 95.2 | 1.364 | 0.0001 | |
| AletheiaModel=TinyLlama-1.1B2026.04 | 26.3 | — | 66.7 | 1.357 | 0.0002 | |
| AletheiaModel=Qwen2.5-0.5B2026.04 | 26.1 | — | 69.4 | 1.353 | 0.0006 | |
| AletheiaModel=Overall2026.04 | 23.1 | — | — | 1.305 | 0.001 | |
| AletheiaModel=Phi-3.5-mini2026.04 | 18.4 | — | 86.2 | 1.225 | 0.0003 | |
| AletheiaModel=Qwen2.5-7B2026.04 | 17.6 | — | 129.7 | 1.214 | 0 | |
| AletheiaModel=Llama-3.1-70B2026.04 | 17.4 | — | 437.5 | 1.211 | 0.0001 | |
| AletheiaModel=Qwen2.5-72B2026.04 | 15.8 | — | 435.9 | 1.187 | 0.0004 | |
| Standard LoRAModel=Qwen2.5-0.5B2026.04 | — | 93.9 | — | — | — | |
| Standard LoRAModel=TinyLlama-1.1B2026.04 | — | 90.6 | — | — | — | |
| Standard LoRAModel=Qwen2.5-1.5B2026.04 | — | 113 | — | — | — | |
| Standard LoRAModel=Phi-3.5-mini2026.04 | — | 105.5 | — | — | — | |
| Standard LoRAModel=Mistral-7B2026.04 | — | 133.1 | — | — | — | |
| Standard LoRAModel=Qwen2.5-7B2026.04 | — | 157.5 | — | — | — | |
| Standard LoRAModel=Llama-3.1-8B2026.04 | — | 129.9 | — | — | — | |
| Standard LoRAModel=Qwen2.5-14B2026.04 | — | 195.1 | — | — | — | |
| Standard LoRAModel=Llama-3.1-70B2026.04 | — | 529.7 | — | — | — | |
| Standard LoRAModel=Qwen2.5-72B2026.04 | — | 517.4 | — | — | — |