List your company

Multi-turn tool-use on AppWorld

63.6Avg@4SeeUPO
2.47418.3432534.212550.08175Dec 11, 2025Dec 20, 2025Dec 30, 2025Jan 8, 2026Jan 18, 2026Jan 27, 2026Feb 6, 2026

Evaluation results

MethodLinks
SeeUPOBackbone=Qwen3-14B, Fine-tuning=RLHF2026.0263.680.7
SeeUPOBackbone=Qwen2.5-14B, Fine-tuning=RLHF2026.0250.8870.18
ReMe (dynamic)Model=Qwen3-32B2025.1242.0263.49
PPOBackbone=Qwen2.5-14B, Fine-tuning=RLHF2026.0240.7957.89
GRPOBackbone=Qwen3-14B, Fine-tuning=RLHF2026.0240.3557.89
GRPOBackbone=Qwen2.5-14B, Fine-tuning=RLHF2026.0235.5349.12
PPOBackbone=Qwen3-14B, Fine-tuning=RLHF2026.0235.0954.39
ReMe (dynamic)Model=Qwen3-14B2025.1234.3252.98
GSPOBackbone=Qwen3-14B, Fine-tuning=RLHF2026.0232.8952.63
ReMe (fixed)Model=Qwen3-32B2025.1231.558.13
A-MemModel=Qwen3-32B2025.1228.1351.19
No MemoryModel=Qwen3-32B2025.1227.2350.59
ReMe (fixed)Model=Qwen3-14B2025.1225.3546.82
ReMe (dynamic)Model=Qwen3-8B2025.1224.742.06
LangMemModel=Qwen3-32B2025.1224.5547.02
GSPOBackbone=Qwen2.5-14B, Fine-tuning=RLHF2026.0224.1238.6
No MemoryModel=Qwen3-14B2025.1222.5741.07
LangMemModel=Qwen3-14B2025.1221.8841.67
Qwen3-14BBackbone=Qwen3-14B, Fine-tuning=None2026.0220.1835.09
A-MemModel=Qwen3-14B2025.1218.9537.7
ReMe (fixed)Model=Qwen3-8B2025.1217.0636.31
No MemoryModel=Qwen3-8B2025.1214.9732.85
A-MemModel=Qwen3-8B2025.1212.9529.76
LangMemModel=Qwen3-8B2025.1211.4626.79
Qwen2.5-14BBackbone=Qwen2.5-14B, Fine-tuning=None2026.024.8257.018