List your company

Multi-hop QA Reasoning on MuSiQue (Degen, Len, AES)

0DegenBase
-26.04149.73325.5501.27Feb 26, 2026

Evaluation results

MethodLinks
BaseModel Scale=Qwen3-0.6B2026.0207310
PromptModel Scale=Qwen3-0.6B2026.020697-0.23
MBT-SModel Scale=Qwen3-4B2026.0204710.85
BaseModel Scale=Qwen3-1.7B2026.0211,1860
GRPOModel Scale=Qwen3-1.7B2026.0211,3790.56
MBT-RModel Scale=Qwen3-1.7B2026.0217281.06
MBT-RModel Scale=Qwen3-4B2026.0217240.65
PromptModel Scale=Qwen3-1.7B2026.0221,1340.05
MBT-SModel Scale=Qwen3-1.7B2026.0225211.37
BaseModel Scale=Qwen3-4B2026.0221,3680
MBT-RModel Scale=Qwen3-0.6B2026.0257502.21
PromptModel Scale=Qwen3-4B2026.0251,324-0.06
MBT-SModel Scale=Qwen3-0.6B2026.0265612.72
RSModel Scale=Qwen3-0.6B2026.0212943-0.04
GRPOModel Scale=Qwen3-0.6B2026.02162,167-0.43
RSModel Scale=Qwen3-1.7B2026.02191,5330.03
GRPOModel Scale=Qwen3-4B2026.02732,433-0.65
RSModel Scale=Qwen3-4B2026.021333,097-1.36
TokenSkipModel Scale=Qwen3-0.6B2026.021743,220-3.7
LIMOProModel Scale=Qwen3-0.6B2026.023865,697-5.26
TokenSkipModel Scale=Qwen3-1.7B2026.024717,437-5.66
LIMOProModel Scale=Qwen3-1.7B2026.025858,393-5.9
LIMOProModel Scale=Qwen3-4B2026.025908,478-6.13
TokenSkipModel Scale=Qwen3-4B2026.026519,932-7.69