ResearchBenchmarksInventory control on InventoryBench 1.0Follow0.538Mean Normalized RewardOR → LLM0.441280.466390.49150.51661Feb 13, 2026Evaluation ResultsMethodMethodLinksMean Normalized Reward95% CIOR → LLMLLM Model=Gemini 3 FlashLLM Model=Gemini 3 Flash2026.020.5380.016LLM → ORLLM Model=Gemini 3 FlashLLM Model=Gemini 3 Flash2026.020.5010.018LLMLLM Model=Gemini 3 FlashLLM Model=Gemini 3 Flash2026.020.4940.018OR2026.020.4450.018