Skip navigation
I asked 6 agents in two conditions (parallel v/s sequential) to create the same tower defence game 👇
0:00 / 1:11

I asked 6 agents in two conditions (parallel v/s sequential) to create the same tower defence game 👇

Paras Chopra

@paraschopra241K followers

Visit on X
55 likesDownloadOriginal post

11,602 views

I asked 6 agents in two conditions (parallel v/s sequential) to create the same tower defence game 👇 LLMs are known to have mode collapse, i.e. if you have multiple parallel runs on the same prompt, you get very similar responses. Is there a way to increase diversity? Yes, it turns out that if you do such multiple runs **sequentially**, instead of in parallel. (Plus for each sequential run, append results from previous runs and instruct the LLM to create something different from before.) I did this quick experiment to explore parallel v/s sequential runs for diversity. Given the same prompt to build a tower defence game, sequential runs had a much higher diversity and innovation v/s parallel runs. Check out this (AI-generated) video for gameplay clips. The sequential cohort created interesting new dynamics and visual style, while parallel ones all collapsed to similar colors and orbital style dynamics.

6 Comments

Comments from the original post on X (Twitter)

  • Legal Lancer India | Hire Freelancers's profile picture

    @freelanceindiaa

    An important distinction: parallel output does not necessarily mean independent thinking. In business workflows, diversity may come less from adding more agents and more from giving them different roles, constraints and evaluation criteria.

  • Vikas(Vik) Malpani| AI for US Real Estate's profile picture

    @vikasmalpani

    Tracks that sequential wins. Parallel copies on one prompt sample the same mode, so N agents give one answer N times. Diversity has to come from each agent seeing the others' work or a different starting context. More heads only help if they aren't reading the same page.

  • Gerald Sterling's profile picture

    @geraldrsterling

    The sequential condition is doing more than adding context. Each prior game becomes a negative prompt for the next one. Apparently agents need peer pressure to develop taste.

  • RXI(E's profile picture

    @PrivateThilak

    excellent idea lol I'm stealing this.

  • Paras Chopra's profile picture

    @paraschopra

    please do!

    1 like

  • urumads🇮🇳's profile picture

    @urumaDS

    Isnt this auto research?