Interactive Decision Making on ALFWorld (test)
96.87Success RateSELAUR
Evaluation results
| Method | Links | ||||||||
|---|---|---|---|---|---|---|---|---|---|
| SELAURBase Model=Qwen2.5-7B2026.02 | 96.87 | — | 100 | 100 | 100 | 100 | 92.59 | 92.59 | |
| GiGPO + PA-MoEType=RL Training, Backbone=Qwen2.5-7B-Instruct2026.02 | 95.3 | — | 87.5 | 100 | 100 | 100 | 100 | 100 | |
| GEPOType=RL Training, Backbone=Qwen2.5-7B-Instruct2026.02 | 94.9 | — | 100 | 80.2 | 98.6 | 95.6 | 94.3 | 86.7 | |
| GiGPOBase Model=Qwen2.5-7B2026.02 | 94.53 | — | 100 | 100 | 100 | 94.44 | 78.94 | 91.66 | |
| GiGPO + PA-MoEType=RL Training, Backbone=Qwen2.5-1.5B-Instruct2026.02 | 93.8 | — | 97.1 | 75 | 95.8 | 100 | 89.5 | 91.7 | |
| COPOBackbone=Qwen2.5-7B-Instruct2026.02 | 92.5 | 1,739.4 | — | — | — | — | — | — | |
| COPOBackbone=Llama3.1-8B-Instruct2026.02 | 91.5 | 811.6 | — | — | — | — | — | — | |
| GiGPOType=RL Training, Backbone=Qwen2.5-7B-Instruct2026.02 | 90.8 | — | 97.7 | 82.7 | 98.8 | 83.7 | 89.3 | 79.2 | |
| GiGPO w/o stdType=RL Training, Backbone=Qwen2.5-7B-Instruct2026.02 | 90.2 | — | 91.8 | 88.6 | 95.9 | 90.2 | 86.5 | 85.2 | |
| COSFT + GiGPOBackbone=Llama3.1-8B-Instruct2026.02 | 89.5 | 6,112 | — | — | — | — | — | — | |
| GEPOType=RL Training, Backbone=Qwen2.5-1.5B-Instruct2026.02 | 89.1 | — | 98.7 | 79.9 | 94.6 | 81.9 | 76.6 | 88.6 | |
| SELAURBase Model=Qwen2.5-1.5B2026.02 | 89.06 | — | 91.43 | 62.5 | 95.83 | 100 | 78.94 | 87.5 | |
| GiGPOBase Model=Qwen2.5-1.5B2026.02 | 88.28 | — | 91.43 | 75 | 100 | 88.89 | 84.21 | 79.16 | |
| COSFT + GiGPOBackbone=Qwen2.5-7B-Instruct2026.02 | 88 | 2,955.5 | — | — | — | — | — | — | |
| GiGPOType=RL Training, Backbone=Qwen2.5-1.5B-Instruct2026.02 | 86.7 | — | 94.4 | 67.5 | 94.8 | 94.4 | 79.8 | 76.4 | |
| GiGPO w/o stdType=RL Training, Backbone=Qwen2.5-1.5B-Instruct2026.02 | 86.1 | — | 96 | 76.5 | 91.8 | 91.3 | 71.7 | 79.5 | |
| RLOOBase Model=Qwen2.5-7B2026.02 | 85.93 | — | 100 | 80 | 100 | 85.71 | 68.75 | 77.78 | |
| COSFT + GRPOBackbone=Llama3.1-8B-Instruct2026.02 | 84 | 3,941.1 | — | — | — | — | — | — | |
| COSFT + GRPOBackbone=Qwen2.5-7B-Instruct2026.02 | 83.5 | 4,994.6 | — | — | — | — | — | — | |
| PPO + PA-MoEType=RL Training, Backbone=Qwen2.5-7B-Instruct2026.02 | 82.6 | — | 93.5 | 67.5 | 93.8 | 91.2 | 82.5 | 71.5 | |
| COSFTexpBackbone=Llama3.1-8B-Instruct2026.02 | 82 | 2,903.2 | — | — | — | — | — | — | |
| AdaptThinkBackbone=Qwen2.5-7B-Instruct2026.02 | 81.5 | 1,372.1 | — | — | — | — | — | — | |
| PPOBase Model=Qwen2.5-1.5B2026.02 | 80.46 | — | 91.42 | 37.5 | 83.33 | 83.33 | 73.68 | 79.16 | |
| PPOType=RL Training, Backbone=Qwen2.5-7B-Instruct2026.02 | 80.4 | — | 92.3 | 64 | 92.5 | 89.5 | 80.3 | 68.8 | |
| GRPO + PA-MoEType=RL Training, Backbone=Qwen2.5-7B-Instruct2026.02 | 80.2 | — | 92.3 | 69.5 | 91.5 | 77.5 | 75 | 67.8 | |
| AdaptThinkBackbone=Llama3.1-8B-Instruct2026.02 | 79.5 | 1,607.2 | — | — | — | — | — | — | |
| RSEARBackbone=Qwen2.5-7B, Evaluation Protocol=+ retry2026.06 | 79.4 | — | — | — | — | — | — | — | |
| COSFTexpBackbone=Qwen2.5-7B-Instruct2026.02 | 78.5 | 3,151 | — | — | — | — | — | — | |
| RLOO + PA-MoEType=RL Training, Backbone=Qwen2.5-7B-Instruct2026.02 | 78.2 | — | 89.2 | 80.5 | 89.5 | 83.5 | 74.2 | 52.5 | |
| RLOOBase Model=Qwen2.5-1.5B2026.02 | 78.12 | — | 90 | 70 | 92 | 71.42 | 68.75 | 50 | |
| GRPOType=RL Training, Backbone=Qwen2.5-7B-Instruct2026.02 | 77.6 | — | 90.8 | 66.1 | 89.3 | 74.7 | 72.5 | 64.7 | |
| ACERBackbone=Qwen2.5-7B, Evaluation Protocol=+ retry2026.06 | 76.7 | — | — | — | — | — | — | — | |
| ReflexionBackbone=Qwen2.5-7B, Evaluation Protocol=+ retry2026.06 | 76.4 | — | — | — | — | — | — | — | |
| GEPARBackbone=Qwen2.5-7B, Evaluation Protocol=+ retry2026.06 | 75.8 | — | — | — | — | — | — | — | |
| RLOOType=RL Training, Backbone=Qwen2.5-7B-Instruct2026.02 | 75.5 | — | 87.6 | 78.2 | 87.3 | 81.3 | 71.9 | 48.9 | |
| GRPOBase Model=Qwen2.5-7B2026.02 | 75 | — | 82.75 | 80 | 92 | 57.14 | 68.75 | 61.11 | |
| RLOO + PA-MoEType=RL Training, Backbone=Qwen2.5-1.5B-Instruct2026.02 | 74.2 | — | 74.3 | 50 | 87.5 | 83.3 | 73.7 | 62.5 | |
| GRPO + PA-MoEType=RL Training, Backbone=Qwen2.5-1.5B-Instruct2026.02 | 74.2 | — | 87.5 | 87.5 | 87.5 | 81.2 | 50 | 45.8 | |
| openai-o3Backbone=Frontier Large Models2026.02 | 74 | 4,290.9 | — | — | — | — | — | — | |
| deepseek-r1-0528Backbone=Frontier Large Models2026.02 | 73 | 6,230.9 | — | — | — | — | — | — | |
| GRPOType=RL Training, Backbone=Qwen2.5-1.5B-Instruct2026.02 | 72.8 | — | 85.3 | 53.7 | 84.5 | 78.2 | 59.7 | 53.5 | |
| PPOBase Model=Qwen2.5-7B2026.02 | 71.09 | — | 72.41 | 70 | 88 | 42.85 | 75 | 61.11 | |
| Dynamic CheatsheetBackbone=Qwen2.5-7B, Evaluation Protocol=Single-pass2026.06 | 70.7 | — | — | — | — | — | — | — | |
| gemini-2.5-pro-0506Backbone=Frontier Large Models2026.02 | 70 | 4,896.5 | — | — | — | — | — | — | |
| RLOOType=RL Training, Backbone=Qwen2.5-1.5B-Instruct2026.02 | 69.7 | — | 88.3 | 52.8 | 71 | 62.8 | 66.4 | 56.9 | |
| ETOBackbone=Qwen2.5-7B-Instruct2026.02 | 69.5 | 540.6 | — | — | — | — | — | — | |
| RSEABackbone=Qwen2.5-7B, Evaluation Protocol=Single-pass2026.06 | 69.3 | — | — | — | — | — | — | — | |
| GRPOBase Model=Qwen2.5-1.5B2026.02 | 67.96 | — | 67.56 | 50 | 80 | 92.85 | 61.9 | 52.63 | |
| ACEBackbone=Qwen2.5-7B, Evaluation Protocol=Single-pass2026.06 | 66.9 | — | — | — | — | — | — | — | |
| AWMBackbone=Qwen2.5-7B, Evaluation Protocol=Single-pass2026.06 | 65.4 | — | — | — | — | — | — | — | |
| ETOBackbone=Llama3.1-8B-Instruct2026.02 | 65 | 730 | — | — | — | — | — | — | |
| ReActBackbone=Qwen2.5-7B, Evaluation Protocol=Single-pass2026.06 | 64.6 | — | — | — | — | — | — | — | |
| GEPABackbone=Qwen2.5-7B, Evaluation Protocol=Single-pass2026.06 | 63.9 | — | — | — | — | — | — | — | |
| SFTBackbone=Qwen2.5-7B-Instruct2026.02 | 62.5 | 494.1 | — | — | — | — | — | — | |
| gpt-4o-2024-08-06Backbone=Frontier Large Models2026.02 | 61.5 | 860.8 | — | — | — | — | — | — | |
| Gemini-2.5-ProType=Prompting, Backbone=Closed-Source Model2026.02 | 60.3 | — | 92.8 | 63.3 | 62.1 | 69 | 25.4 | 58.7 | |
| COSFTBackbone=Qwen2.5-7B-Instruct2026.02 | 57 | 3,912.7 | — | — | — | — | — | — | |
| PPO + PA-MoEType=RL Training, Backbone=Qwen2.5-1.5B-Instruct2026.02 | 56.2 | — | 88.6 | 37.5 | 50 | 55.6 | 36.8 | 37.5 | |
| COSFTBackbone=Llama3.1-8B-Instruct2026.02 | 55.5 | 3,546.6 | — | — | — | — | — | — | |
| SFTBackbone=Llama3.1-8B-Instruct2026.02 | 54.5 | 632 | — | — | — | — | — | — | |
| PPOType=RL Training, Backbone=Qwen2.5-1.5B-Instruct2026.02 | 54.4 | — | 64.8 | 40.5 | 57.1 | 60.6 | 46.4 | 47.4 | |
| deepseek-v3-0324Backbone=Frontier Large Models2026.02 | 52 | 1,171.2 | — | — | — | — | — | — | |
| SkillLens2026.05 | 51.31 | — | 65.71 | 76.92 | 37.04 | 50 | 24 | 54.17 | |
| ReActBackbone=Qwen2.5-7B-Instruct2026.02 | 50.5 | 666.4 | — | — | — | — | — | — | |
| GPT-4oType=Prompting, Backbone=Closed-Source Model2026.02 | 48 | — | 75.3 | 60.8 | 31.2 | 56.7 | 21.6 | 49.8 | |
| AutoSkill2026.05 | 47.14 | — | 62.86 | 76.92 | 37.04 | 37.5 | 24 | 50 | |
| EvoSkill2026.05 | 47.14 | — | 65.71 | 76.92 | 40.74 | 31.25 | 20 | 50 | |
| Baseline2026.05 | 45 | — | 65.71 | 76.92 | 25.93 | 43.75 | 16 | 50 | |
| ReflexionType=Prompting, Backbone=Qwen2.5-7B-Instruct2026.02 | 42.7 | — | 62 | 41.6 | 44.9 | 30.9 | 36.3 | 23.8 | |
| ReflexionBackbone=Qwen2.5-7B-Instruct2026.02 | 42.5 | 2,340.8 | — | — | — | — | — | — | |
| ReActType=Prompting, Backbone=Qwen2.5-7B-Instruct2026.02 | 31.2 | — | 48.5 | 35.4 | 34.3 | 13.2 | 18.2 | 17.6 | |
| ReflexionType=Prompting, Backbone=Qwen2.5-1.5B-Instruct2026.02 | 21.8 | — | 35.3 | 22.2 | 21.7 | 13.6 | 19.4 | 3.7 | |
| ReflexionBackbone=Llama3.1-8B-Instruct2026.02 | 18.5 | 3,645.8 | — | — | — | — | — | — | |
| Qwen2.5Type=Prompting, Backbone=Qwen2.5-7B-Instruct2026.02 | 14.8 | — | 33.4 | 21.6 | 19.3 | 6.9 | 2.8 | 3.2 | |
| ReActBackbone=Llama3.1-8B-Instruct2026.02 | 13 | 959.4 | — | — | — | — | — | — | |
| ReActType=Prompting, Backbone=Qwen2.5-1.5B-Instruct2026.02 | 12.8 | — | 17.4 | 20.5 | 15.7 | 6.2 | 7.7 | 2 | |
| promptingBase Model=Qwen2.5-7B2026.02 | 11.71 | — | 28 | 25 | 12 | 0 | 3.7 | 0 | |
| ReAct + R1-DistillBackbone=Qwen2.5-7B-Instruct2026.02 | 4.5 | 33,620.5 | — | — | — | — | — | — | |
| Qwen2.5Type=Prompting, Backbone=Qwen2.5-1.5B-Instruct2026.02 | 4.1 | — | 5.9 | 5.5 | 3.3 | 9.7 | 4.2 | 0 | |
| promptingBase Model=Qwen2.5-1.5B2026.02 | 3.12 | — | 10.71 | 0 | 0 | 7.1 | 0 | 0 | |
| ReAct + R1-DistillBackbone=Llama3.1-8B-Instruct2026.02 | 2.5 | 21,720.8 | — | — | — | — | — | — |