List your company

Logical Reasoning on K&K tasks

2.13Perplexity (3ppl)Vanilla
-1.759624.495250.7577.0048May 25, 2026

Evaluation results

MethodLinks
VanillaModel=Qwen3-1.7B-Base2026.052.131.50.370.750.1200.81
VanillaModel=Qwen3-4B-Base2026.059.257.634.6331.750.634.48
TTRLModel=Qwen3-1.7B-Base2026.0510.634.871.881.751.251.253.6
ProbModel=Qwen3-1.7B-Base2026.0511.632.885.122.750.6303.6
EntropyModel=Qwen3-1.7B-Base2026.0513.122.885.122.750.6304.08
RandomModel=Qwen3-1.7B-Base2026.0518.251054.250.8806.4
VanillaModel=Phi4-mini-instruct2026.0521.881686.3731.759.5
TTRLModel=Qwen3-8B-Base2026.0524.8810.376.758.12249.35
VanillaModel=Qwen3-8B-Base2026.0525.2517.513.888.755.53.6212.42
RandomModel=Qwen3-8B-Base2026.0525.75—8.2510.632.374.1310.54
OracleModel=Qwen3-1.7B-Base2026.0527.3820.8715.7515.254514.71
EntropyModel=Qwen3-4B-Base2026.0528.2516.7514.1210.534.1312.79
ProbModel=Qwen3-4B-Base2026.0528.371811.8711.122.624.8712.81
GTModel=Qwen3-1.7B-Base2026.0540.2530.131920.876.758.1220.85
ProbModel=Qwen3-8B-Base2026.0541.1230.3811.214.255417.79
CAREModel=Qwen3-1.7B-Base2026.0541.3730.517.7518.636.511.3421.02
RandomModel=Phi4-mini-instruct2026.0545.531.871918.513.129.1322.85
EntropyModel=Qwen3-8B-Base2026.05464132.6232141329.77
TTRLModel=Phi4-mini-instruct2026.0546.8833.6321.6320.3713.510.1324.35
ProbModel=Phi4-mini-instruct2026.0547.3733.8824.2520.513.3812.2525.27
EntropyModel=Phi4-mini-instruct2026.0548.253526.2523.2514.2510.8826.31
TTRLModel=Qwen3-4B-Base2026.0550.2535.3728.1321.8812.879.2526.29
OracleModel=Phi4-mini-instruct2026.0554.2544.532.8833.523.517.2534.31
RandomModel=Qwen3-4B-Base2026.0556.634328.7524.3813.8810.8829.58
CAREModel=Phi4-mini-instruct2026.0564.554.62—3625.522.2540.04
GTModel=Phi4-mini-instruct2026.0564.7554.1241.7538.1227.882141.27
OracleModel=Qwen3-4B-Base2026.058987.2574.8763.3859.2554.3771.42
OracleModel=Qwen3-8B-Base2026.059593.586.7578.574.2569.6382.94
CAREModel=Qwen3-4B-Base2026.0597.7596.2589.583.1376.7568.6285.33
GTModel=Qwen3-8B-Base2026.059997.87978988.7584.6292.71
GTModel=Qwen3-4B-Base2026.0599.3798.1295.2588.3885.1274.7590.17
CAREModel=Qwen3-8B-Base2026.0599.379893.259186.7577.3890.96