Hallucination
A confident-sounding model output that is false or unsupported by the available evidence.
TL;DR
Confident but incorrect model output; mitigated by RAG citations.
Quick facts
- Category
- AI & knowledge management
- Product tie-in
- KobiGPT RAG platform
- Related
- See compare and tools pages
- Locale
- TR and EN site
Why teams choose KobiGPT
- Understand terms before evaluating vendors.
- Link concepts to KobiGPT features (RAG, Kobi Kredi).
- Share glossary links with procurement and legal.
- Explore assistant use cases next.
Product facts
- Ücretsiz plan
- 100 doküman · 2 departman · 120 Kobi/ay(PLAN_CONFIG)
- Starter
- 1000 doküman · 5 departman · 1000 Kobi/ay(PLAN_CONFIG)
- Pro
- 12500 doküman · 25 departman · 12500 Kobi/ay(PLAN_CONFIG)
- Retrieved context
- Answers are built from source passages and scoped document retrieval(apps/api/src/rag/qdrant.service.ts → searchHybrid)
What hallucination looks like
A model can produce a fluent, persuasive sentence as part of text generation; fluency does not make it true. An invented date, a policy clause absent from the source, or a recommendation that fills in missing conditions are examples.
The danger is not poor prose but excessive confidence. When a corporate user sees a citation, they should still check that the source passage actually supports the claim.
How RAG reduces the risk
RAG anchors the model to passages retrieved from company documents and can show the passage behind an answer. That makes the output more auditable than an unrestricted guess from general model knowledge.
Wrong retrieval, stale documents, and contradictory policies can still produce a wrong answer. A citation exposes the risk; it does not remove the need for verification.
A safer operating pattern
Keep assistant scope focused, separate superseded files, and have a person read the source passage for consequential decisions. Saying that the context is insufficient is more useful than filling a gap with an invented sentence.
In legal, finance, HR, or security contexts, position the assistant as a faster path to the relevant procedure, not as the final decision-maker. Approval should remain with an authorised person.
Applying What hallucination looks like in a controlled workflow
A useful way to evaluate hallucination is to follow one real question from the source document to the final answer. Record which file was selected, what context reached the model, and what a reviewer would need to verify. This turns a definition into an operational check and makes the result comparable across teams.
The same check should include ownership and change management. Decide who updates the relevant documents, how an outdated result is reported, and which access boundary applies. KobiGPT can provide the assistant and the cited document context, but the organisation still owns the source material, permissions, and the decision made from the answer.
FAQ
What is Hallucination in practice?
Confident but incorrect model output; mitigated by RAG citations.
Does KobiGPT use this?
See product docs and feature pages for implementation details.
More reading?
Visit our blog and FAQ.
Accuracy disclaimer?
Educational content; verify for compliance decisions.
Does a citation eliminate hallucination?
No. Retrieval can select the wrong passage; the citation lets you test whether the claim is actually supported.
What should an assistant do when the document has no answer?
It should say the context is insufficient instead of guessing.
Comparison
| Feature | KobiGPT | Alternative |
|---|---|---|
| SME focus | Yes | N/A |
| Citations | When using RAG | N/A |
| Glossary depth | Growing | N/A |
| Tools | Interactive | N/A |