ResearchBenchmarksQuestion Answering on HotPotQA-Vanilla In-DomainFollow0.54AURCBASE0.40480.43990.4750.5101Jul 3, 2026Evaluation ResultsMethodMethodLinksAURCBASEModel=Llama-3.1-8BModel=Llama-3.1-8B2026.070.54RLVRModel=Llama-3.1-8BModel=Llama-3.1-8B2026.070.46RLCRModel=Llama-3.1-8BModel=Llama-3.1-8B2026.070.43RLSRModel=Llama-3.1-8BModel=Llama-3.1-8B2026.070.41