List your company

Question Answering on ARC (test) with Generation Metrics

93Pass@1GRPO
74.07278.98683.988.814Feb 19, 2026

Evaluation results

MethodLinks
GRPOBackbone=Qwen3-4B-Instruct2026.0293161.948.04
DRGRPOBackbone=Qwen3-4B-Instruct2026.0292.6268.4712.34
GRPOBackbone=Qwen2.5-7B-Instruct2026.0290.4309.314.72
DRGRPOBackbone=Qwen2.5-7B-Instruct2026.0290.4226.1111.51
WS-GRPOBackbone=Qwen3-4B-Instruct2026.0288.620.232
WS-GRPOBackbone=Qwen2.5-7B-Instruct2026.0287.9162
DRGRPOBackbone=Llama-3.1-8B-Instruct2026.0284.181
WS-GRPOBackbone=Llama-3.1-8B-Instruct2026.0282.4202
GRPOBackbone=Llama-3.1-8B-Instruct2026.0281.613.051
GRPOBackbone=Llama-3.2-3B-Instruct2026.0279.4137.25.07
DRGRPOBackbone=Llama-3.2-3B-Instruct2026.0278.297.914.22
WS-GRPOBackbone=Llama-3.2-3B-Instruct2026.0274.8142.04