The Video Assistant · Lesson 6 — The Phantom Call ← Course

The phantom call — hallucinations.

Sometimes the model hands you a confident, fluent answer that is simply made up. That is a hallucination — and there are ways to make it much rarer.

A hallucination is when the model states something false as if it were fact — confidently, fluently, often with convincing detail. It is not lying; it does not know it is wrong. It just produced the continuation its training made most likely.

Why does it happen? Recall Lesson 1: the model only ever predicts the next word, and it always produces something fluent. It has no reliable internal alarm for “I don't know” — and no default way to express one. When it lacks the real answer, the most natural-sounding continuation is often a natural-sounding invention. Gaps are the commonest cause, but a leading question or a strong pattern can pull an invention out of a model that “knew” better. The confident tone is not evidence of truth — it reads the same whether the model knows or is guessing.

Picture a video assistant referee expected to make a call on every incident. He did not see the contact, but his instinct fires anyway and the whistle goes — no deceit, just a reflex producing a call where there was no evidence. That is the phantom call. The model does the same, without the choice: fluent continuation is all it produces, so a gap comes out as fluent invention rather than “I couldn't tell.”

The fixes are not about trusting harder. You can (1) give it permission to say “I don't know”, (2) bring it the actual evidence to read from instead of asking from memory — the big one — and (3) ask it to cite a source so you can check. Lowering temperature (Lesson 5) removes one small source — the odd unlucky token — but not a confident fabrication; grounding it in evidence is the biggest lever.

Ask the VAR something it cannot know, watch it invent an answer — then add the safeguards one by one and watch the phantom call give way.

A question with no record behind it.
Step 1 of 4
you ask the VAR
safeguards
✓allowed to say “not sure”
✓given the real record to read
✓must cite its source
↓
the VAR answers

The safeguards are just how you call it.

Each safeguard is a change to the request, not to the model. The biggest one — handing over the record — is a whole technique of its own, coming next. The record goes where everything else goes: into the context window from Lesson 2, paying the same budget. 


      

    


    

      
It cannot not answer — so a gap comes out as an answer.

      

        A model always produces fluent text, so a gap in its knowledge — or a leading question — comes out as a
        confident, plausible fabrication — not a blank and not a warning. The tone tells you nothing about whether it is true — and neither does the model's
        own confidence: a made-up answer can carry just as high a probability as a real one.
      

      

        The fix is in the setup, not in trusting harder: let it abstain, bring it the
        real evidence to read from (retrieval — the next lesson), and ask it to cite so
        you can check. And for anything that matters, verify.
      

      

        
Quick check

        
The VAR describes a foul that never happened — fluently and with confidence. This is…

        

          
          
          
        

        

      

      

        Next: the footage bank →
        
      

    


  


  

    

      ▸ Go deeper — why it happens and what actually helps optional
      

        
“Confident” tells you nothing

        

          The fluency and certainty in the wording are generated the same way whether the
          model knows the answer or is inventing it — there is no honesty dial that leaks
          into the prose. Even the model's internal probability can be miscalibrated, so a
          made-up answer can read every bit as assured as a correct one. Never treat tone
          as evidence.
        

        
Grounding is the biggest lever

        

          The most reliable systems don't ask the model to recall facts from memory at all.
          They fetch the relevant documents and have the model answer from those, which is the whole of the next lesson. A model reading off
          the record hallucinates far less than one recalling from training.
        

        
Give it an exit

        

          Simply telling the model it is allowed to say “I don't know” measurably cuts down
          invented answers. Models often fabricate because the prompt implicitly demands an
          answer and leaves no acceptable way to decline. Offer one.
        

        
It can still be wrong — so verify

        

          None of these make a model perfect; they lower the rate, not the possibility. For
          anything that matters, check the cited source yourself. A citation can be invented too; its value is that a fake one is easy to catch. Measuring how often a model
          goes wrong, and on what, is its own discipline, and a later lesson of its own.