Severity Classification on Smart Contract Audit Dataset benchmark leaderboard · SOTA2 Research

Severity Classification on Smart Contract Audit Dataset

61.97PrecisionOurs (rsLoRA + Voting)

Evaluation Results

Method	Links
Ours (rsLoRA + Voting)2026.06		61.97	57.45	58.77	64.58
LoRA (with voting)2026.06		59.61	52.8	50.81	63.19
CodeLlama-34B2026.06		54.72	47.76	37.75	36.81
CodeLlama-13B2026.06		52.78	36.39	19.05	26.39
CodeGemma-7B2026.06		40.78	47.38	43.32	61.11
Qwen2.5-14B2026.06		39.23	47.11	42.6	59.72
CodeLlama-34B2026.06		39.03	40.64	36.9	58.33
Qwen2.5-32B2026.06		39.01	46.13	41.94	59.03
Codestral-22B2026.06		38.8	47.82	42.79	59.03
DeepSeek-7B2026.06		36.78	37.7	33.31	56.25
CodeLlama-13B2026.06		35.35	36.72	31.97	55.56
LoRA (w/o voting)2026.06		35.29	35.59	29.81	55.56
rsLoRA (w/o voting)2026.06		33.71	39.12	35.21	54.86
Qwen2.5-32B2026.06		31.98	45.96	34.78	46.53
Qwen2.5-14B2026.06		25.21	38.31	25.58	34.72
DeepSeek-7B2026.06		23.65	30.71	24.54	33.33
Codestral-22B2026.06		18.52	33.33	23.81	55.56
DeepSeek-33B2026.06		18.52	33.33	23.81	55.56
DeepSeek-33B2026.06		18.52	33.33	23.81	55.56
CodeGemma-7B2026.06		17.31	32.21	14.01	23.61