Qwen3 LLMs Trained for Creativity Using Word Association Game

Researchers have developed a novel method called Reinforcement Learning with Verifiable Rewards (RLVR) to train Large Language Models (LLMs) for creativity, bypassing subjective human judgment. They applied this technique to Qwen3 models of varying sizes (1.7B, 4B, and 8B parameters) using the word-association game Codenames. The study found that larger models, like the 8B version, demonstrated improved creativity across multiple benchmarks with only minor reasoning degradation, while smaller models prioritized reasoning precision over creative association. AI

IMPACT Introduces a scalable method for training LLMs in creative tasks, potentially improving their utility in content generation and problem-solving.

RANK_REASON The cluster contains an academic paper detailing a new training methodology for LLMs and evaluating specific model versions. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.CL →

• Codenames
• Large Language Models (LLMs)
• Qwen3
• Reinforcement Learning with Verifiable Rewards (RLVR)

• paper
• model release

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →