Models from Team TüLK at SemEval-2026 Task 1 (Humor Generation): GRPO-trained Qwen2.5-7B policies + joke-rater reward models.
-
KonradBRG/Qwen2.5-7B-Instruct-Jokester
Text Generation • 8B • Updated • 376 • 1 -
KonradBRG/Qwen2.5-7B-Instruct-Jokester-English
Text Generation • 8B • Updated • 363 -
KonradBRG/Qwen2.5-7B-Instruct-Jokester-Spanish
Text Generation • 8B • Updated • 368 -
KonradBRG/Qwen2.5-7B-Instruct-Jokester-Chinese
Text Generation • 8B • Updated • 374 • 1