Hello everyone! I recently developed a simple LLM-based test generator. It can generate tests about any topic. If you are interested, you can check a few examples that were generated by it here: [link redacted]<p>It can make mistakes time to time but overall it's still usable. To try it out, just visit the website: [link redacted]. You will need an API key from an LLM provider however.<p>You may wonder why not just use the LLM directly. That could work fine too actually but this tool puts it neatly on a webpage with formatting like LaTeX, answer choices if requested, etc. You can convert the page into PDF too (by printing to PDF).
Hello fellow Edtech programmer,<p>TLDR: > 0.0% hallucination = mistrust from people<p>We're in the same boat, I can give a quick relay of the feedback i was getting.<p>Generally about test/education generation with model is that they're unreliable because of hallucinations. people hate that very much as you may be aware, in my case I was asking for payment on top so it made the feedback even harsher.<p>One way I see that you can strike gold and maybe even successfully monetise this is if you allow the user to provide urls that you scrape, then adjust your prompts for a successful (Retrieval augmented generation) RAG. This way people will trust the tool and may consider paying if the features solve a headache for them (this is the direction im considering for my APP).