About PYQPrep and how its answer keys are audited
PYQPrep is a question bank of previous year questions from Indian postgraduate medical entrance exams, for NEET PG preparation, split by subject and topic, with timed tests marked the way the exam marks them and a diagnosis of where a student is losing marks.
Every answer key here has been re-solved blind, by a solver that had to commit to its own answer before it was allowed to read the stored key, and a key is changed only when a second, independent round reaches the same conclusion.
Who runs it
PYQPrep is run by one person, in India. There is no editorial board here, no panel of reviewers and no bylined author, and this page names nobody on purpose: a name on a medical page is a claim about a person rather than about the work. What is offered instead is the process below. Every step of it is recorded in files committed to this project, and every figure on this page is read from one of them.
Where the questions come from
Most questions here come from previous-year papers of Indian postgraduate medical entrance and licensing exams; where a paper and year are recorded the question carries them. The image-based questions are original vignettes written around an openly licensed figure, and were never asked in any exam.
Everything on this site is practice for NEET PG (National Eligibility cum Entrance Test - Postgraduate). The ones that carry a paper and a year came from that paper itself, not from anyone here.
Each image question is credited under its image to the paper the picture came from, and none is ever given an exam or a year, because there is none to give.
Where a paper and a year are not recorded against a question, it carries neither: a year is never inferred from the wording of a question, and a question that merely mentions an exam is not treated as having come from it. That rule is the reason the badges can be believed at all.
How the answer key is audited
The rule the whole audit is built on is that one opinion never rewrites a medical question. A solver is given a question and its options and has to write down its own answer, and commit it, before it is allowed to read the stored key. A pass at high confidence changes nothing but a timestamp. Anything else - a proposed fix, a retirement, or any verdict the solver was not confident in - needs a second, independent round that reaches the same result, down to the same option. Whatever is left over, where there was only one round or where two rounds disagreed, goes to a human queue and rewrites nothing.
The bank that existed before that gate was put through it in full: 8,074 questions in 160 batches, each batch applied only on the terms above.
One weakness in that audit was found afterwards and closed. The 617 questions that had passed only because the solver, unsure, deferred to the stored key were all re-solved blind. The 61 that came back as confident disputes went to a second solver who saw neither the key nor the first opinion. Where both independently named the same different answer, the key changed: 47 times. Another 2 questions that neither could answer as written were retired, and the remaining 12 disputes were left exactly as they were, because honest doubt is not evidence.
The questions built around an image were held to the same rule when they were written: each was solved blind by a second model that saw the question and its image but not the key, and every disagreement was settled against the paper the image came from. A later pass graded 1,067 of them a second time against a written style guide, and that pass could not touch a key at all. It rewrote wording only: every proposed edit was independently re-reviewed before it could be applied, a proposal the second reader rejected was dropped, and the apply step refused outright any change that moved the correct option.
None of that closes the loop on its own, so a student can. Any question can be reported from the question itself, saying what is wrong with it. Every report is resolved by hand as upheld or rejected, and the student who filed it sees that outcome on their own profile page.
What this site does not publish
PYQPrep publishes past-year questions and answer keys. It does not publish written explanations. There is no explanation data behind the bank, and writing one for every question would put medical prose on the site that nobody had checked - the opposite of everything above. When a question needs more than the key, a textbook is the right place to go.
Nothing on this site is medical advice, and none of it is a substitute for a textbook or a teacher. It is exam preparation, and the claim it makes is narrower and checkable: this is where each question came from, and this is what was done to be sure of the key.
Where to go next
The bank is the point of all of this. Practise a subject and see what it looks like, or write in if something here does not add up.