The Base State Hypothesis: A Falsification Protocol for a Substrate Capable of Hosting the Standard Model, Audited Against Contemporary Quantum Platforms

In the paper's words

We formalize the question of whether a single, explicitly constructible quantum object — a "Base State" — could serve as a substrate on which the structure of the Standard Model is hosted, rather than the Standard Model being assembled from particle-like primitives on a qubit-gate substrate. A Base State is defined as a normalized state (or density operator) in a finite-dimensional Hilbert space, equipped with a declared readout map and a declared symmetry group, such that every empirically accessible structure arises by preparation, measurement, and composition. We audit feasibility using the only quantitative hardware data in our source corpus: the Tianyan-287 superconducting processor (105 qubits; single-qubit gate fidelity 99.90\%, two-qubit gate fidelity 99.56\%, readout fidelity 98.7\%). [...]

The abstract, verbatim, and its opening sentence is the claim. Every line below is quoted from the text.

Numbers it reports

1. With full arithmetic we derive: per-operation error rates of $1.0\times10^{-3}$, $4.4\times10^{-3}$, and $1.3\times10^{-2}$; a two-qubit serial budget of at most 157 gates at $50\%$ aggregate fidelity (154 with one final readout); a full-register readout ceiling of $F_{\mathrm{ro}}\approx0.2532$; a representative 105-qubit circuit fidelity of $F_{\mathrm{circ}}\approx0.1467$; and a parameter count of $2^{106}-2\approx8.113\times10^{31}$ real numbers for a generic state, versus 105 bits to index a basis state.

   from Abstract

2. Full-register readout of 105 qubits at $F_r = 0.987$ caps verification fidelity at $F_{\mathrm{ro}} \approx 0.2532$ (Section 4.4).

   from 5. Results

3. This work describes the Tianyan Quantum Cloud Platform, offering cloud services demonstrating quantum advantage capabilities with a Zuchongzhi 3.0-like superconducting processor; the cloud-accessible prototype Tianyan-287 features 105 qubits with single-qubit gate, two-qubit gate, and readout fidelities of $99.90\%$, $99.56\%$, and $98.7\%$, respectively.

   from 2. Background and Related Work

4. (This scenario assumes readout of all 105 qubits; the single-readout scenario of Section 4.3 is the contrasting convention — see Appendix A, D2.)

   from 4.4 Readout-only ceiling for full-register verification

5. Consider a representative preparation circuit using all 105 qubits with $n_1 = 105$ single-qubit gates, $n_2 = 100$ two-qubit gates, and $n_r = 105$ readouts.

   from 4.5 Representative full-register circuit

How it is done

1. With full arithmetic we derive: per-operation error rates of $1.0\times10^{-3}$, $4.4\times10^{-3}$, and $1.3\times10^{-2}$; a two-qubit serial budget of at most 157 gates at $50\%$ aggregate fidelity (154 with one final readout); a full-register readout ceiling of $F_{\mathrm{ro}}\approx0.2532$; a representative 105-qubit circuit fidelity of $F_{\mathrm{circ}}\approx0.1467$; and a parameter count of $2^{106}-2\approx8.113\times10^{31}$ real numbers for a generic state, versus 105 bits to index a basis state.

   from Abstract

2. Third, it proposes a falsification protocol in the spirit of the automata-learning literature, where a model is recovered from noisy, incomplete data by evidence-driven merging rather than by optimizing a numeric target [1], and in the spirit of the Falsification Pledge of the Manifesto for Honest Computation [12].

   from 1. Introduction

3. This work presents an interactive version of an evidence-driven state-merging (EDSM) algorithm for learning variants of finite state automata, framing learning as recovering or reverse-engineering the model that generated noisy, incomplete, or imperfectly sampled data, with domain expertise injected by a human in the loop.

   from 2. Background and Related Work

4. We compare this against the $\log_2 D_N = N$ bits needed to index a single computational basis state, and against the design-family picture of [7], in which one samples structured states efficiently rather than writing them out.

   from 3.3 Specifiability accounting

What would falsify it

1. Following the model-recovery stance of [1] and the Falsification Pledge of [12], the protocol is: Stage P (Preparation) : prepare a candidate $|\psi_0\rangle$ on the full register; record circuit depth and gate counts.

   from 3.4 Falsification protocol

2. Stage V (Verification) : run a statistics test with threshold $\theta = 0.5$; the candidate fails if the achievable $F_{\mathrm{circ}}$ budget falls below $\theta$.

   from 3.4 Falsification protocol

3. We propose a three-stage falsification protocol, state explicit falsifiers, and conclude that the Base State hypothesis is currently unfalsified but unconfirmed: the obstruction is quantitative, not a no-go in principle.

   from Abstract

Where the claim stops

1. The fidelity model of Section 3.2 assumes independent, multiplicative errors and ignores crosstalk, leakage, idling decay, and state-preparation error.

   from 6. Discussion

2. How symmetry constrains state designs remains open per [7]; whether the categorical language of [8] can express the gauge structure $SU(3)\times SU(2)\times U(1)$ without particle ontology is untested here; and the excited-state machinery of [5] has not been adapted to Base State dynamics.

   from 6. Discussion

3. This is directly relevant methodologically: inferring a Base State from measurement data is a model-recovery problem under noise, not a numeric optimization, and the ESDM pattern — merge states when evidence supports it, consult an oracle when it does not — is a template for how a candidate Base State's equivalence classes of measurement outcomes could be inferred. [2] Tianyan: Cloud services with quantum advantage (arXiv:2512.10504v2).

   from 2. Background and Related Work

Where to read

1. Abstract210 words · 1 min
2. 1. Introduction300 words · 1 min
3. 2. Background and Related Work1,030 words · 4 min
4. 3. Methods602 words · 3 min
5. 4. Analysis1,034 words · 4 min
6. 5. Results388 words · 2 min
7. 6. Discussion348 words · 2 min
8. 7. Conclusion123 words · 1 min
9. Appendix A. Divergence report181 words · 1 min
10. Appendix B. Claim attribution198 words · 1 min