List your company

Summarization on WikiHow (Reward & Baseline Comparison)

0.39Reward ScoreRL (Proposed)
0.102960.177480.2520.32652Mar 19, 2026

Evaluation results

MethodLinks
RL (Proposed)Model=Granite-3.3-2B2026.030.3917.175.6922.88
StaticModel=Granite-3.3-2B2026.030.369———
GreedyModel=Granite-3.3-2B2026.030.333———
RL (Proposed)Model=Qwen-2.5-0.5B2026.030.20478.9545.71-43.6
StaticModel=Qwen-2.5-0.5B2026.030.14———
GreedyModel=Qwen-2.5-0.5B2026.030.114———