List your company
ResearchFormerly WizWand

Downstream Repair on GSM8K trace-level 60/20/20 (test)

88.8Repair AccuracyFull trace
69.24874.32479.484.476May 28, 2026
Updated 5mo ago

Evaluation results

MethodLinks
Full traceRepair model=Gemma 4 E4B IT, Alpha=0.05, Input Strategy=Full trace2026.0588.8——
Whole-trace abst.Repair model=Gemma 4 E4B IT, Alpha=0.05, Input Strategy=Whole-trace abst.2026.0588.6——
CROPRepair model=Gemma 4 E4B IT, Alpha=0.05, Input Strategy=CROP prefix2026.0588.3-0.91-1.53
CROPRepair model=Qwen2.5-7B-Instruct, Alpha=0.05, Input Strategy=CROP prefix2026.0587.60.950.09
CROPRepair model=DeepSeek-R1-0528-Qwen3-8B, Alpha=0.05, Input Strategy=CROP prefix2026.0587.41.830.89
Whole-trace abst.Repair model=Qwen2.5-7B-Instruct, Alpha=0.05, Input Strategy=Whole-trace abst.2026.0586.3——
Full traceRepair model=Qwen2.5-7B-Instruct, Alpha=0.05, Input Strategy=Full trace2026.0585.1——
Full traceRepair model=DeepSeek-R1-0528-Qwen3-8B, Alpha=0.05, Input Strategy=Full trace2026.0584.9——
Whole-trace abst.Repair model=DeepSeek-R1-0528-Qwen3-8B, Alpha=0.05, Input Strategy=Whole-trace abst.2026.0584.6——
Question onlyRepair model=Gemma 4 E4B IT, Alpha=0.05, Input Strategy=Question only2026.0582.3——
Question onlyRepair model=Qwen2.5-7B-Instruct, Alpha=0.05, Input Strategy=Question only2026.0580.5——
Whole-trace abst.Repair model=Llama3.1-8B, Alpha=0.05, Input Strategy=Whole-trace abst.2026.0574.6——
CROPRepair model=Llama3.1-8B, Alpha=0.05, Input Strategy=CROP prefix2026.0573.9-1.18-2.02
Full traceRepair model=Llama3.1-8B, Alpha=0.05, Input Strategy=Full trace2026.0573.3——
Question onlyRepair model=Llama3.1-8B, Alpha=0.05, Input Strategy=Question only2026.0571.1——
Question onlyRepair model=DeepSeek-R1-0528-Qwen3-8B, Alpha=0.05, Input Strategy=Question only2026.0570——