ResearchBenchmarksCode Reasoning on LiveCodeBench v6 (Acc avg@32)Follow73.8Accuracy avg@32IOP-GSPO60.07263.63667.270.764Apr 19, 2026Evaluation ResultsMethodMethodLinksAccuracy avg@32IOP-GSPOModel Architecture=Qwe...Model Architecture=Qwen3-Next2026.0473.8GSPOModel Architecture=Qwe...Model Architecture=Qwen3-Next2026.0469.6BaseModel Architecture=Qwe...Model Architecture=Qwen3-Next2026.0468.7IOP-GSPOModel Architecture=Qwe...Model Architecture=Qwen3-32B2026.0467.8GSPOModel Architecture=Qwe...Model Architecture=Qwen3-32B2026.0461.8BaseModel Architecture=Qwe...Model Architecture=Qwen3-32B2026.0460.6