List your company

Reasoning on SuperGPQA (Accuracy, Length)

7,262LengthBase
1,061.522,671.264,2815,890.74May 8, 2026

Evaluation results

MethodLinks
BaseBackbone=Qwen3-8B2026.057,26244.2
BaseBackbone=Qwen3-4B2026.056,85837.2
BaseBackbone=DSQW-7B2026.056,23830
+GRPOBackbone=DSQW-7B2026.055,45231.8
+ICRBackbone=DSQW-7B2026.054,63533.2
+GRPOBackbone=Qwen3-4B2026.054,14341
+GRPOBackbone=Qwen3-8B2026.054,02242.6
+LC-R1Backbone=DSQW-7B2026.053,83232.6
+ICRBackbone=Qwen3-8B2026.053,49844
+LP-FBackbone=Qwen3-8B2026.053,39241.6
+LP-FBackbone=DSQW-7B2026.053,39033.4
+LP-G (Kimi)Backbone=DSQW-7B2026.053,27731.8
+LP-FBackbone=Qwen3-4B2026.053,18938.2
+ICRBackbone=Qwen3-4B2026.053,08241.2
+LC-R1Backbone=Qwen3-8B2026.053,07940.4
+ShorterBetterBackbone=DSQW-7B2026.052,96731.6
+ICR /w LP-FBackbone=DSQW-7B2026.052,95735.2
+LC-R1Backbone=Qwen3-4B2026.052,88436.2
+ICR /w LP-FBackbone=Qwen3-4B2026.052,39239.6
+ICR /w LP-FBackbone=Qwen3-8B2026.052,31241.6
+LP-G (Kimi)Backbone=Qwen3-8B2026.052,28039.4
+ShorterBetterBackbone=Qwen3-8B2026.052,26940.8
+ShorterBetterBackbone=Qwen3-4B2026.052,00934.8
+LP-G (Kimi)Backbone=Qwen3-4B2026.051,30037.2