Lipschitz Ball Verification on Qwen 7B-Instruct LoRA 2.5 (train)
79Ball Acceptance RateLipschitz ball verifier
Evaluation results
| Method | Links | |||||||||
|---|---|---|---|---|---|---|---|---|---|---|
| Lipschitz ball verifierBackbone model=Qwen2.5-7B-Instruct, Fine-tuning method=LoRA, LoRA rank=4, Target modules=Wq and Wv, Number of trainable parameters=1,261,568, Oracle Evaluation=10-prompt2026.03 | 79 | 42 | 0 | 42 | 9.64 | 90.59 | 0.011 | 2.583 | 1 | |
| Lipschitz ball verifierBackbone model=Qwen2.5-7B-Instruct, Fine-tuning method=LoRA, LoRA rank=4, Target modules=Wq and Wv, Number of trainable parameters=1,261,568, Oracle Evaluation=50-prompt2026.03 | 75.5 | 49 | 20 | — | 69.1 | — | — | 2.093 | 0.98 |