Fabricated claim not found in referenced link

jonathan-s

In the following thread https://assistant.kagi.com/share/69e4aa31-e8e6-4a96-9ad7-d00490965907

I'm asking about actuation force of mice and ask for a top 10. The top result there reference a link that does not contain the stated claim. I'd very much wish for a second round of fact checking such that when a given link is referenced we check that a claimed fact is indeed correct.

In this case it turns out the actuation force for the top result was completely fabricated.

The desired functionality is that when you have a link reference any claim from that link reference should indeed be true, and not fabricated like above.

Pleonasm

The root cause of the problem may be that Kagi Search, as used by Kagi Assistant, does not (AFAIK) have a robust mechanism for "grounding," as explained by Microsoft:

Grounding an AI–generated answer introduces a fundamentally different constraint: the system is no longer just pointing to information, it is using it. The goal shifts from “fetch the best documents” to “fetch the best information to synthesize into a reliable, verifiable answer.”
Instead of just ranking pages, the index must help an AI system determine which specific information can responsibly support an answer. The unit of value shifts from documents to groundable information – discrete, supportable facts with clear provenance.... Grounding consequently emphasizes high-quality source identification and attribution so users (and downstream systems) can verify what was used and follow the evidence when needed.

It is challenging to see how Kagi can successfully compete in the long-term in the absence of incorporating such technological advancements.

jonathan-s

I believe the usage of reference already provides a certain amount of grounding, having a separate LLM call to reason about whether it can find that in the source material would likely ground it even further.