List your company

Structured tax-form completion on TaxCalcBench TY24 (train)

90Train-A ScoreDPA-GRPO
51.5261.5171.581.49May 8, 2026

Evaluation results

MethodLinks
DPA-GRPOBackbone=Qwen3-8B2026.05906858
Zero-shotBackbone=Qwen3-8B2026.05844242
GRPOBackbone=Qwen3-8B2026.05846332
DAPOBackbone=Qwen3-8B2026.05846353
GSPOBackbone=Qwen3-8B2026.05846858
GRPOBackbone=Qwen3-4B2026.05744732
DAPOBackbone=Qwen3-4B2026.05743721
GSPOBackbone=Qwen3-4B2026.05743721
DPA-GRPOBackbone=Qwen3-4B2026.05745837
Zero-shotBackbone=Qwen3-4B2026.05533226