Adjudicate one item you got wrong, and find the misreading
Takes a single wrong answer and returns the deciding sentence, the False-versus-Not-Given logic, the exact feature you misread, and a 10-second in-exam test.
I got this item wrong. Adjudicate it. Passage extract: [PASTE THE RELEVANT PARAGRAPHS] Statement: [ITEM] Official answer: [ANSWER] · My answer: [YOUR ANSWER] 1. Quote the exact sentence(s) in the passage that determine the answer. If none exists, say so — the item may be flawed. 2. Explain the decision in terms of the F/NG boundary: does the passage CONTRADICT the statement (False), or merely FAIL TO ADDRESS it (Not Given)? 3. Name the specific feature I probably misread: a hedge, a scope word (some/all/ most), a comparison never made, a cause never claimed, or my own world knowledge. 4. Give me a 10-second test I could have run in the exam to get this right. 5. Give me 3 more items on the same passage that test the same boundary, with answers.
Best on Claude · ChatGPT · Gemini.
What to fill in
| Token | What to paste |
|---|---|
| [PASTE THE RELEVANT PARAGRAPHS] | The two or three paragraphs around the answer, not the whole passage. |
| [ITEM] | The statement exactly as it was written in the question paper. |
| [ANSWER] | The official answer from the key: True, False or Not Given. |
| [YOUR ANSWER] | What you actually put. |
What you should get back
The deciding quote, a False-versus-Not-Given explanation, the specific feature you misread, a fast test you could have run in the exam, and three fresh items on the same boundary.
What it actually returned
Tested on Gemini 2.5 Flash, API, August 2026, 2026-08-06. Trimmed — this is an excerpt of a longer response, not the whole thing.
You've hit a very common and tricky pitfall in True/False/Not Given questions! Let's break it down. 1. Quote the exact sentence(s) in the passage that determine the answer. "In parts of the ancient Near East, donkeys remained the pack animal of choice long after horses were available, partly because they tolerated heat and poor fodder better. Cost mattered too: a horse required grain in a way a donkey did not…" 2. Explain the decision in terms of the F/NG boundary: The statement claims that "Donkeys were preferred to horses across the ancient Near East because they were cheaper to feed." — The passage confirms that donkeys were preferred in "parts of the ancient Near East". This is a crucial difference in scope. "Across" implies the entire region, while "in parts" means some of the region, but not necessarily all.
Our verdict on this run
Past the opening exclamation, this was the most useful run of the fifteen. It quoted the deciding sentences, then isolated the exact word doing the work — "across" against the passage's "in parts" — which is precisely the scope-word misreading that produces most False-versus-Not-Given errors. Naming the feature rather than restating the answer is what makes the lesson transfer. It did have the official answer in front of it, so this is a run where the model reasoned toward a key it already knew; withhold the key if you want to test whether it reaches the same place unaided.
How this prompt goes wrong
The model rationalises the official answer instead of adjudicating it. Because you supply the key, it treats that answer as ground truth and constructs a justification for it — including when the practice item is genuinely broken, which is common in unofficial material. Point 1 is the guard: if it cannot quote a sentence that settles the claim, take the "flawed item" possibility seriously rather than accepting an explanation built on a sentence that does not say what it claims. Withhold the official answer and run it blind if you want an independent read.
Tips
- False-versus-Not-Given confusion is the highest-frequency Reading error, and it is a logic problem rather than a vocabulary one.
- Point 4 is the part that transfers to the exam. Collect those 10-second tests in one note.
- Run this only on items you got wrong. Adjudicating correct answers teaches nothing and burns study time.
Related prompts
- ReadingGenerate True/False/Not Given items with a named trap eachForces the model to choose one of seven trap types before writing each statement, then prove the answer with the deciding sentence from the passage.
- ReadingBuild a paraphrase drill where no content word repeatsTurns any passage into 12 paraphrase-matching items with three deliberate near-misses, and a hidden key that names each distortion.
- Writing Task 2Make the model argue your essay is half a band worseAsks the model to build the strongest case against the band it just gave you, then arbitrate honestly between the two.
- SpeakingTurn a recording of yourself into countable speaking dataCounts your words per minute, filled pauses, restarts, repeated words and structure range from a transcript — and refuses to guess a band.
Common questions
What if the practice item itself is wrong?
That happens more often than most candidates assume in unofficial material. Point 1 is designed to surface it: if no sentence in the passage settles the claim, the answer key is asserting something the text does not support.
Should I tell it the official answer?
It gives a more focused explanation if you do, but it will also rationalise a bad key. For an independent read, withhold both answers, get its verdict, then compare.
How do I stop making this mistake?
Almost every False/Not Given error comes from answering with world knowledge instead of the text. The habit that fixes it is mechanical: before answering, point at the sentence that decides it. If you cannot point, it is Not Given.