Your observation about repetitive answers is valid, and it stems from how Large Action Models (LAMs) are designed to operate. As previously explained in the forum, this issue often arises because you’re interacting with different agents or instances of the LAM, rather than a single, persistent AI agent (like perplexity, claude, gpt, etc.).
Each time you engage with a LAM-based system, it generates responses based on its current input and context. If you’re asking similar questions across multiple agents or sessions, the system lacks memory of previous interactions unless explicitly designed to retain it. This is why you might receive the same response repeatedly—it doesn’t “know” what it has already told you unless continuity is built into the system.
To address this:
- Use a single agent consistently: If possible, interact with a single instance of the LAM that has memory capabilities or a shared context. This can help reduce redundancy.
- Reframe your queries: Instead of asking for “something new” or “you have not told me before”, provide additional context or specify what kind of novelty you’re looking for. This helps guide the LAM toward generating more unique responses.
- Explore systems with advanced memory: Some LAM implementations are paired with memory modules that allow them to recall past interactions. These systems are better suited for ongoing conversations without repetition.
The key takeaway is that LAMs are powerful but inherently stateless unless designed otherwise. By changing how you interact with them or exploring more advanced configurations, you can unlock their potential for more dynamic and varied responses.
Note: if you want to continue in this way, saying something similar at “you have not told me before”, you must at least save in the notes all previous jokes that it said…
I hope this can help you to open your mind…
I haven’t tried this, but try saving the joke it always tells you as the first joke as a note and then the next times try telling r1 to use the joke note to tell you a new joke that it has never told you before… in my opinion this is how it starts to work and start the ecosystem that will allow r1 to continue saving as a note all the new jokes it will tell you and, therefore, will be different…
How to do?
Say it to save a note called “smile” with …here the joke that r1 repeats…
How to recall?
Use the note called “smile” TO TELL ME A JOKE YOU HAVE NOT TOLD ME BEFORE
Try it and let us know…