lingjiao chen on X: "A surprising finding: a larger number of LLM calls can incur worse performance of compound AI systems! Why and what is the desired number of LLM calls? We initialize the study of scaling properties of compound AI systems both theoretically and empirically: https://t.co/MWmmgosENz https://t.co/w9u42nJ3av" / X

To view keyboard shortcuts, press question mark View keyboard shortcuts Home Explore Notifications Messages Grok Lists Bookmarks Communities Premium Profile More Post Amos You @amooseyou Post See new posts Conversation lingjiao chen @ChenLingjiao A surprising finding: a larger number of LLM calls can incur worse performance of compound AI systems! Why and what is the desired number of LLM calls? We initialize the study of scaling properties of compound AI systems both theoretically and empirically: https://arxiv.org/pdf/2403.02419.pdf… 5:03 PM · Mar 5, 2024 · 10.7K Views 1 12 25 19 Post your reply Reply lingjiao chen @ChenLingjiao · 9h Focusing on one-layer Voting Inference Systems, we find (i) that item difficulty plays an important role, and (ii) that predicting the scaling trend and optimal number of LLM calls is possible. 1 3 277 lingjiao chen @ChenLingjiao · 8h It was very fun working on this project with Jared Quincy Davis, @BorisHanin , @pbailis , Ion Stoica, @matei_zahari

saved by

• Amos You