How statement and proof provenance work
The first chip identifies the source of the statement or construction; the second identifies the source of its local proof or verification.
- Literature-sourced: the exact statement appears in a cited source; only wording and notation differ.
- AI-adapted: a semantically identical restatement of literature-sourced material, modulo indexing, notation, and boundary cases adopted by the library.
- AI-generated: a genuinely novel statement formulated by AI, with no source for the claim itself.
These labels describe origin, not correctness: citations and verification chips remain separate evidence.
The Polynomial Hierarchy and Relativization: Examples and Counterexamples
1 · Prerequisites
- Boolean Circuits and Nonuniform Complexity
- Construction of the Natural Numbers
- Construction of the Real Numbers via Cauchy Sequences
- Construction of the Real Numbers via Dedekind Cuts
- Countability and Uncountability
- Finite Counting, Factorials and Binomial Coefficients
- Formal Languages, Encodings, and Decision Problems
- Foundations of the Real Numbers for Analysis
- Linear Recurrences and Rational Generating Functions
- Logarithmic Space, NL, and Reachability
- P, NP, coNP, and Polynomial Reductions
- Relations, Functions, and Quotients
- Resource Bounds and Machine Invariance
- Sequences and Limits
- The Polynomial Hierarchy and Relativization
- The ZFC Axioms and the Basic Set Constructions
- Time and Space Hierarchy Theorems
- Turing Machines, Configurations, and Computation
2 · Summary
These examples distinguish quantifier order, total oracle access, and promise correctness. A separation in one oracle world does not justify a conclusion valid in every oracle world.
3 · Logical flowchart
4 · Definitions, theorems and proofs
None yet.
5 · Examples, counterexamples and false statements
Np and conp are the first levels
Example
Formula satisfiability and tautology exhibit the existential and universal first levels. For a finite simple graph and integer , the language defined by belongs to ; no completeness claim is made for it. Malformed inputs and or are rejected.
Facts & Assumptions
Given: the objects and hypotheses in the statement above.
and . (Np and conp are the first levels).
For every fixed , truth of prenex Boolean formulas with alternating blocks beginning existentially is -complete under polynomial-time many-one reductions. The universally starting version is -complete. The matrix is a general Boolean formula, empty blocks are allowed, and malformed encodings are rejected. (Bounded alternation qbf is level complete).
Verification
A formula is satisfiable exactly when some assignment makes its polynomial-time evaluation true, and is a tautology exactly when all assignments do. These are first-level predicates, with completeness for the corresponding bounded-QBF encodings. A formula with no variables has one assignment.
For valid , let test that an -bit indicator represents an independent set of cardinality . Then the exact condition is . Reversing the two independent blocks gives the same truth value: in the reverse direction fix any to obtain a witness , and every must fail . If both conditions hold one witness works for every . At the empty set witnesses ; at , is always false. This also covers the graph with no vertices. A deterministic validity test in the matrix rejects all malformed and out-of-range inputs for either polarity.
Relativization alone cannot resolve p versus np
Example
A TQBF oracle absorbs polynomial nondeterministic computation into PSPACE. A finite diagonal stage with clock and can defeat a machine on , provided all earlier committed words have length less than eight. This finite stage is not a complete separating oracle.
Facts & Assumptions
Given: the objects and hypotheses in the statement above.
Neither an argument for nor an argument for can be valid unchanged relative to every total oracle. (Relativization alone cannot resolve p versus np).
Verification
A branch has at most polynomially many queries, all of polynomial length. Replaying the finite branch tree and deciding each TQBF query in reusable polynomial space gives the absorption; one many-one TQBF reduction supplies the converse PSPACE inclusion into deterministic oracle time.
At there are candidate words and at most queries. Preserve earlier commitments and answer every new query NO. At least length-eight words remain unqueried. An accepting run is defeated by freezing the entire length to NO; a rejecting run by setting one unqueried word YES and freezing the rest NO. Also freeze every queried word of other lengths. Further stages are required to defeat other programs. Opposite full oracle worlds, not this one finite count alone, yield the relativization obstruction.
Relativized separations prove unrelativized separations
Statement refuted
The oracle-independent transfer schema refuted in Relativized separations prove unrelativized separations fails: a separating oracle cannot force separation in every other oracle world.
Facts & Assumptions
Given: the objects and hypotheses in the statement above.
The refutation supplies total oracles and with and , respectively. These are counterinstances to the universal transfer rule; no conclusion about the empty oracle follows. (Relativized separations prove unrelativized separations).
Counterexample
Use the separating oracle and the equality oracle in the refutation of the schema. They satisfy and , respectively. Thus the premise holds while one instance of the universal conclusion fails.
The empty oracle is a third specified world. The exhibited failure at establishes invalidity of universal transfer, but gives no equality or inequality for the empty oracle. Hence this counterexample does not resolve the ordinary P-versus-NP problem.
Promise oracle off promise answers
Example
For the target promise , , a caller which accepts its sole promised YES input exactly when the target answers YES to query is not a valid oracle promise reduction.
Facts & Assumptions
Given: the objects and hypotheses in the statement above.
For promise problems and in the stated convention, a polynomial-time many-one promise reduction is a total polynomial-time function satisfying and . There is no condition on outside . A polynomial-time oracle promise reduction is a polynomially clocked deterministic oracle machine which solves for every total language satisfying and . This uses binary, consistent membership completions: an off-promise word may have either bit, but repeated queries to that word receive the same bit. The clock is uniform over all completions. (Promise preserving reduction).
A total membership oracle in the stated convention fixes the answer on every query word. A promise target in the stated convention describes a collection of total completions. Correctness of a promise reduction must hold for each completion, including its arbitrary answers outside the target promise. A promised input to the caller does not by itself guarantee that the caller's queries satisfy the target promise. (Oracle and promise conventions are distinct).
Verification
Both and are total membership completions respecting and . They disagree on the off-promise word . Their unspecified words are simply NO membership answers.
Take the source pair . On input the caller rejects with and accepts with . Universal correctness over completions therefore fails at a promised YES input, despite the caller's constant running time. No source-NO obligation is needed for this failure.
Sources
- Arora–Barak, Computational Complexity, 2007 draft; Examples5.2–5.3 and5.9, pp91–94.
- Arora–Barak, Computational Complexity, 2007 draft; §3.5 proof of Theorem3.9, pp71–72.
- Arora–Barak, Computational Complexity, 2007 draft; §3.5, pp71–72.
- Goldreich, On Promise Problems; §1.2 oracle-reduction convention, p5; finite illustration.