ResearchBenchmarksScientific Reasoning on GPQA D (Avg@4)Follow34.7Avg@4GRPO28.66830.23431.833.366May 8, 2026Evaluation ResultsMethodMethodLinksAvg@4GRPOBackbone=Llama3.2-3B-I...Backbone=Llama3.2-3B-Instruct, Domain Type=In-domain2026.0534.7Dr.GRPOBackbone=Llama3.2-3B-I...Backbone=Llama3.2-3B-Instruct, Domain Type=In-domain2026.0534.3PRIMEBackbone=Llama3.2-3B-I...Backbone=Llama3.2-3B-Instruct, Domain Type=In-domain2026.0533.1OriginalBackbone=Llama3.2-3B-I...Backbone=Llama3.2-3B-Instruct, Domain Type=In-domain2026.0528.9