Embodied Instruction Following on AlfWorld (Specific Success Rates)
99.3Average Success RateGemini-3-Pro
Evaluation Results
| Method | Links | ||||||||
|---|---|---|---|---|---|---|---|---|---|
| Gemini-3-ProBackbone=Closed-Source Model2026.05 | 99.3 | 100 | 96.8 | 100 | 100 | — | 100 | 100 | |
| WALL-E 2.0Model Category=Language model2026.04 | 98 | 100 | 100 | 96 | 100 | 100 | 94 | — | |
| DAPO+AEMBackbone=Qwen2.5-7B-Instruct2026.05 | 96.6 | 99 | 100 | 96.3 | 95.2 | — | 93.2 | 91.7 | |
| DAPOBackbone=Qwen2.5-7B-Instruct2026.05 | 96.1 | 100 | 100 | 94.7 | 90.3 | — | 94.3 | 96.3 | |
| ReCAPAModel Category=Language model2026.04 | 96 | 100 | 97 | 94 | 95 | 96 | 94 | — | |
| RAFAModel Category=Language model2026.04 | 95 | 100 | 97 | 91 | 95 | 100 | 82 | — | |
| WALL-E 1.0Model Category=Language model2026.04 | 95 | 100 | 97 | 100 | 86 | 85 | 100 | — | |
| DAPO+AEMBackbone=Qwen2.5-1.5B-Instruct2026.05 | 94.5 | 97.3 | 98.8 | 98.4 | 90.9 | — | 89.5 | 90.3 | |
| AdaPlannerModel Category=Language model2026.04 | 91 | 100 | 100 | 89 | 100 | 97 | 47 | — | |
| GPT-5.2-ProBackbone=Closed-Source Model2026.05 | 88.8 | 100 | 100 | 61.3 | 87 | — | 100 | 100 | |
| DAPOBackbone=Qwen2.5-1.5B-Instruct2026.05 | 88.5 | 100 | 90.6 | 91.3 | 86.6 | — | 82.9 | 70.3 | |
| ReflexionModel Category=Language model2026.04 | 86 | 92 | 94 | 70 | 81 | 90 | 88 | — | |
| DELTAMEMLLM=DeepSeek-V4-flash, Decoding Temperature=0, Prompt Framework=ReAct, Node Embeddings=E5-base-v2, Learning Protocol=Online learning, in-context demonstration=nicl = 12026.06 | 85 | — | — | — | — | — | — | — | |
| GRPO+AEMBackbone=Qwen2.5-7B-Instruct2026.05 | 84.4 | 98.9 | 89.4 | 84.1 | 79.5 | — | 65.7 | 78.6 | |
| GSPO+AEMBackbone=Qwen2.5-7B-Instruct2026.05 | 83.4 | 88.9 | 92.6 | 85.2 | 84.8 | — | 78.3 | 56.8 | |
| EMMAModel Category=Vision-language model2026.04 | 82 | 71 | 94 | 85 | 83 | 88 | 67 | — | |
| GSPOBackbone=Qwen2.5-7B-Instruct2026.05 | 80.7 | 95.1 | 73.9 | 80 | 79.8 | — | 69.7 | 66.9 | |
| PPOBackbone=Qwen2.5-7B-Instruct2026.05 | 80.4 | 92.3 | 92.5 | 89.5 | 80.3 | — | 68.8 | 64 | |
| SynapseLLM=DeepSeek-V4-flash, Decoding Temperature=0, Prompt Framework=ReAct, Node Embeddings=E5-base-v2, Learning Protocol=Online learning, in-context demonstration=nicl = 12026.06 | 79.29 | — | — | — | — | — | — | — | |
| GRPOBackbone=Qwen2.5-7B-Instruct2026.05 | 78.7 | 91.3 | 79.9 | 76.9 | 75.2 | — | 44.3 | 91.5 | |
| AutoGenModel Category=Language model2026.04 | 77 | 92 | 74 | 78 | 86 | 83 | 41 | — | |
| GRPO+AEMBackbone=Qwen2.5-1.5B-Instruct2026.05 | 76.8 | 88.6 | 76.4 | 60.9 | 76.7 | — | 69.9 | 67.6 | |
| DEPSModel Category=Language model2026.04 | 76 | 93 | 50 | 80 | 100 | 100 | 0 | — | |
| RBankLLM=DeepSeek-V4-flash, Decoding Temperature=0, Prompt Framework=ReAct, Node Embeddings=E5-base-v2, Learning Protocol=Online learning, in-context demonstration=nicl = 12026.06 | 74.29 | — | — | — | — | — | — | — | |
| ReActModel Category=Language model2026.04 | 74 | 79 | 54 | 96 | 85 | 83 | 51 | — | |
| GSPO+AEMBackbone=Qwen2.5-1.5B-Instruct2026.05 | 71.9 | 75.5 | 78.1 | 75 | 70.2 | — | 46.7 | 56.5 | |
| No MemoryLLM=DeepSeek-V4-flash, Decoding Temperature=0, Prompt Framework=ReAct, Node Embeddings=E5-base-v2, Learning Protocol=Zero-shot, in-context demonstration=nicl = 12026.06 | 70.71 | — | — | — | — | — | — | — | |
| GRPOBackbone=Qwen2.5-1.5B-Instruct2026.05 | 68 | 78.2 | 70.5 | 72 | 75 | — | 39.2 | 49.9 | |
| GSPOBackbone=Qwen2.5-1.5B-Instruct2026.05 | 66.7 | 75.4 | 64.6 | 70 | 74.6 | — | 30 | 54.2 | |
| AWMLLM=DeepSeek-V4-flash, Decoding Temperature=0, Prompt Framework=ReAct, Node Embeddings=E5-base-v2, Learning Protocol=Online learning, in-context demonstration=nicl = 12026.06 | 64.29 | — | — | — | — | — | — | — | |
| PPOBackbone=Qwen2.5-1.5B-Instruct2026.05 | 54.4 | 64.8 | 57.1 | 60.6 | 46.4 | — | 47.4 | 40.5 | |
| ReActBackbone=Qwen2.5-7B-Instruct2026.05 | 31.2 | 48.5 | 34.3 | 13.2 | 18.2 | — | 17.6 | 35.4 | |
| BUTLERModel Category=Language model2026.04 | 26 | 31 | 41 | 60 | 27 | 12 | 29 | — | |
| InstructBLIPModel Category=Vision-language model2026.04 | 22 | 50 | 26 | 23 | 6 | 17 | 0 | — | |
| MiniGPT-4Model Category=Vision-language model2026.04 | 16 | 4 | 0 | 19 | 17 | 67 | 6 | — | |
| LLaMA-AdapterModel Category=Vision-language model2026.04 | 13 | 17 | 10 | 27 | 22 | 0 | 0 | — | |
| ReActBackbone=Qwen2.5-1.5B-Instruct2026.05 | 12.8 | 17.4 | 15.7 | 6.2 | 7.7 | — | 2 | 20.5 | |
| BLIP-2Model Category=Vision-language model2026.04 | 4 | 0 | 6 | 4 | 11 | 6 | 0 | — |