List your company

End-task performance evaluation on SRA-Bench

71.6TheoremQA ScoreOracle Skill
30.83241.4165262.584Apr 27, 2026

Evaluation results

MethodLinks
Oracle SkillModel=Qwen3-32B2026.0471.686.651.270.983.555.267.2
Oracle SkillModel=Qwen3-4B2026.0469.785.147.170.973.545.561.6
LLM SelectionModel=Qwen3-32B2026.0469.381.144.165.982.548.162.4
Oracle SkillModel=Llama-3.3-70B2026.0468.581.148.768.679.754.764.4
Progressive DisclosureModel=Qwen3-235B2026.0468.181.136.972.277.150.359.9
Full-Skill InjectionModel=Qwen3-32B2026.0468.170.536.270.459.54954.3
LLM SelectionModel=Qwen3-235B2026.0466.879.545.473.582.549.762.8
Oracle SkillModel=Mistral3.1-24B2026.0466.378.648.266.878.65463.2
Oracle SkillModel=Qwen3-235B2026.0465.986.652.375.884.55767.5
Progressive DisclosureModel=Qwen3-32B2026.0464.974.735.558.771.144.755.3
Full-Skill InjectionModel=Qwen3-235B2026.0464.472.83868.666.250.456.2
Full-Skill InjectionModel=Qwen3-4B2026.046469.130.268.636.141.545.5
LLM SelectionModel=Qwen3-4B2026.0463.1703962.365.739.753.3
LLM SelectionModel=Llama-3.3-70B2026.0462.769.743.358.779.450.459.2
LLM DirectModel=Qwen3-235B2026.0461.676.136.466.458.246.753.3
Full-Skill InjectionModel=Llama-3.3-70B2026.0460.166.13561.959.552.952.7
LLM DirectModel=Llama-3.3-70B2026.0459.660.530.956.553.94547.8
LLM SelectionModel=Mistral3.1-24B2026.045966.742.655.676.346.856.6
Full-Skill InjectionModel=Mistral3.1-24B2026.0458.857.234.459.654.645.648.5
LLM DirectModel=Qwen3-32B2026.0457.475.93565.553.943.950.8
Progressive DisclosureModel=Mistral3.1-24B2026.0454.855.827.154.762.642.846.7
Progressive DisclosureModel=Llama-3.3-70B2026.0453.365.329.954.359.545.748.5
Progressive DisclosureModel=Qwen3-4B2026.0452.765.726.760.14537.743.2
LLM DirectModel=Qwen3-4B2026.0450.77525.656.12236.438.8
Oracle SkillModel=Llama-3.1-8B2026.0449.469.523.340.86235.244.5
LLM DirectModel=Mistral3.1-24B2026.0449.356.129.252.549.641.143.4
Progressive DisclosureModel=Llama-3.1-8B2026.0436.95016.425.159.631.436.3
Full-Skill InjectionModel=Llama-3.1-8B2026.0436.558.313.627.836.734.232.7
LLM SelectionModel=Llama-3.1-8B2026.0435.153.819.424.75732.137
LLM DirectModel=Llama-3.1-8B2026.0432.454.616.722.426.932.329.8