List your company

Reasoning on BRUMO25 (avg@k, Token reduction)

78.75Avg@k ScoreType IV (STOP)
27.613240.889154.16567.4409Apr 17, 2026

Evaluation results

MethodLinks
Type IV (STOP)Model=DS-Qwen-3-8B2026.0478.75277.578.98
Type IModel=DS-Qwen-3-8B2026.0476.25284.878.43
Type IIModel=DS-Qwen-3-8B2026.0475.83394.970.1
No pruning (Baseline)Model=DS-Qwen-3-8B2026.04751,320—
Type IIIModel=DS-Qwen-3-8B2026.0474.58280.278.78
Type IV (STOP)Model=GPT-OSS-20B2026.0470183.668.11
Type IModel=GPT-OSS-20B2026.0469.17187.667.41
No pruning (Baseline)Model=GPT-OSS-20B2026.0468.02575.6—
Type IIIModel=GPT-OSS-20B2026.0467.9217968.9
Type IIModel=GPT-OSS-20B2026.0466.25298.848.09
Type IV (STOP)Model=DS-Qwen-2.5-7B2026.0456.67190.271.03
Type IIIModel=DS-Qwen-2.5-7B2026.0452.92194.570.38
Type IIModel=DS-Qwen-2.5-7B2026.0451.25312.152.46
No pruning (Baseline)Model=DS-Qwen-2.5-7B2026.0450.99656.6—
Type IModel=DS-Qwen-2.5-7B2026.0450199.369.64
Type IV (STOP)Model=DS-Qwen-2.5-1.5B2026.0433.75204.473.61
Type IIModel=DS-Qwen-2.5-1.5B2026.0431.67325.657.96
Type IIIModel=DS-Qwen-2.5-1.5B2026.0431.67209.772.93
No pruning (Baseline)Model=DS-Qwen-2.5-1.5B2026.0430.99774.6—
Type IModel=DS-Qwen-2.5-1.5B2026.0429.58212.872.53