#passat1 — Public Fediverse posts
Live and recent posts from across the Fediverse tagged #passat1, aggregated by home.social.
-
New research from Tsinghua shows that reasoning‑augmented LLMs solve tasks with fewer calls but don’t surpass raw capability. The study compares chain‑of‑thought prompting, RL‑based RLVR, and pass@1 metrics, highlighting efficiency gains for open‑source models. Worth a read for anyone tracking LLM benchmarks. #LLM #ChainOfThought #RLVR #PassAt1
🔗 https://aidailypost.com/news/study-finds-reasoning-llms-are-more-efficient-not-more-capable