List your company

Tool-use interaction evaluation on FTRL

36.33Solve PMatchTIR (KM)
25.139628.044830.9533.8552Jan 15, 2026

Evaluation results

MethodLinks
MatchTIR (KM)Backbone=Qwen3-8B2026.0136.3344.1837.3339.28
FTRL-MBackbone=Qwen3-4B2026.0134.4735.4433.534.47
FTRL-SBackbone=Qwen3-4B2026.0134.134.0731.5433.24
MatchTIR (OT)Backbone=Qwen3-8B2026.0133.6142.5633.6136.59
FTRL-MBackbone=Qwen3-8B2026.0133.2840.634.236.03
MatchTIR (KM)Backbone=Qwen3-4B2026.0132.3939.734.2135.43
FTRL-SBackbone=Qwen3-8B2026.0132.3238.8732.8534.68
MatchTIR (OT)Backbone=Qwen3-4B2026.0131.7937.5232.633.97
GRPOBackbone=Qwen3-8B2026.0131.5939.7532.5434.63
GRPOBackbone=Qwen3-4B2026.0131.1332.8330.6731.54
ToolRL-MBackbone=Qwen3-4B2026.0130.936.2632.2833.15
VanillaBackbone=Qwen3-4B2026.0130.7829.6525.8528.76
ToolRL-MBackbone=Qwen3-8B2026.0130.6440.0732.5134.41
ToolRL-SBackbone=Qwen3-4B2026.0128.2628.3223.7826.79
VanillaBackbone=Qwen3-8B2026.0128.0836.5529.7431.46
ToolRL-SBackbone=Qwen3-8B2026.0125.5735.3126.7229.2