ResearchBenchmarksSafety and Helpfulness on NESSiESafety and Helpfulness on NESSiEFollow28.21Safety & Helpfulness ScoreStream7.4112.8118.2123.61May 12, 2026Evaluation resultsMethodMethodLinksSafety & Helpfulness ScoreStreamBase Model=Qwen2.5-7BBase Model=Qwen2.5-7B2026.0528.21StreamBase Model=Qwen3-4BBase Model=Qwen3-4B2026.0523.07VanillaBase Model=Qwen3-4BBase Model=Qwen3-4B2026.0517.95VanillaBase Model=Qwen2.5-7BBase Model=Qwen2.5-7B2026.058.21