List your company

Logical Reasoning on Web of Lies

59.2PerformanceDART
18.6429.1739.750.23May 26, 2026

Evaluation results

MethodLinks
DARTBase Model=Qwen3-0.6B2026.0559.2
STaRBase Model=Qwen3-0.6B2026.0553.4
TESSYBase Model=Qwen3-0.6B2026.0552.4
BaseBase Model=Qwen3-0.6B2026.0550.5
GRPOBase Model=Qwen3-0.6B2026.0549.5
DARTBase Model=Qwen2.5-0.5B-Instruct2026.0546.9
BaseBase Model=Qwen2.5-0.5B-Instruct2026.0541.1
STaRBase Model=Qwen2.5-0.5B-Instruct2026.0540.4
GRPOBase Model=Qwen2.5-0.5B-Instruct2026.0539.7
Original-SFTBase Model=Qwen2.5-0.5B-Instruct2026.0529.8
TESSYBase Model=Qwen2.5-0.5B-Instruct2026.0525.8
Original-SFTBase Model=Qwen3-0.6B2026.0520.2