Tool Use on IHEval Overall Tool Use v1 (All)
69.9Average AccuracyQwen3-4B-it-CoT
Evaluation Results
| Method | Links | |
|---|---|---|
| Qwen3-4B-it-CoTBackbone=Qwen3-4B-it, Inference Mode=Chain-of-Thought2026.04 | 69.9 | |
| Qwen3-4B-it-NSBackbone=Qwen3-4B-it, Inference Mode=Neuro-Symbolic2026.04 | 68.3 | |
| Qwen3-4B-it-NSHA-DPOBackbone=Qwen3-4B-it, Alignment=NSHA-DPO2026.04 | 67.4 | |
| Qwen3-4B-it-NSHA-HCALBackbone=Qwen3-4B-it, Alignment=NSHA-HCAL2026.04 | 67 | |
| Qwen3-4B-it-NSHA-SFTBackbone=Qwen3-4B-it, Alignment=NSHA-SFT2026.04 | 60.5 | |
| Qwen3-4B-itBackbone=Qwen3-4B-it2026.04 | 51.9 | |
| Llama3.1-8B-NSBackbone=Llama3.1-8B, Inference Mode=Neuro-Symbolic2026.04 | 39 | |
| Llama3.1-8B-NSHA-HCALBackbone=Llama3.1-8B, Alignment=NSHA-HCAL2026.04 | 38.7 | |
| Llama3.1-8BBackbone=Llama3.1-8B2026.04 | 32.1 | |
| Llama3.1-8B-CoTBackbone=Llama3.1-8B, Inference Mode=Chain-of-Thought2026.04 | 31.1 | |
| Llama3.1-8B-NSHA-DPOBackbone=Llama3.1-8B, Alignment=NSHA-DPO2026.04 | 29.6 | |
| Llama3.1-8B-NSHA-SFTBackbone=Llama3.1-8B, Alignment=NSHA-SFT2026.04 | 9 |