List your company

Reasoning on MMLU-Pro (Accuracy and Length)

2,993Output LengthBase
468.921,124.211,779.52,434.79May 8, 2026

Evaluation results

MethodLinks
BaseBackbone=DSQW-7B2026.052,99356.61
BaseBackbone=Qwen3-8B2026.052,63176.07
+GRPOBackbone=DSQW-7B2026.052,51557.32
BaseBackbone=Qwen3-4B2026.052,31272.86
+ICRBackbone=DSQW-7B2026.052,13659.82
+LP-G (Kimi)Backbone=DSQW-7B2026.051,87758.39
+LC-R1Backbone=DSQW-7B2026.051,69357.67
+LP-FBackbone=DSQW-7B2026.051,68857.85
+ShorterBetterBackbone=DSQW-7B2026.051,67657.85
+ICR /w LP-FBackbone=DSQW-7B2026.051,66659.28
+GRPOBackbone=Qwen3-8B2026.051,48679.1
+GRPOBackbone=Qwen3-4B2026.051,26872.32
+ICRBackbone=Qwen3-8B2026.051,19579.46
+LP-FBackbone=Qwen3-4B2026.051,00071.6
+LC-R1Backbone=Qwen3-8B2026.0594676.42
+ICRBackbone=Qwen3-4B2026.0583073.39
+LP-FBackbone=Qwen3-8B2026.0581179.28
+LC-R1Backbone=Qwen3-4B2026.0580572.14
+ICR /w LP-FBackbone=Qwen3-8B2026.0575478.03
+ICR /w LP-FBackbone=Qwen3-4B2026.0575273.39
+LP-G (Kimi)Backbone=Qwen3-8B2026.0575175.17
+ShorterBetterBackbone=Qwen3-8B2026.0572576.78
+LP-G (Kimi)Backbone=Qwen3-4B2026.0572071.25
+ShorterBetterBackbone=Qwen3-4B2026.0556671.25