arXiv:2609.27129
Quantum Soundness of a Total-Degree Line-versus-Point Test
Honghao Fu, Nengkun Yu, Tianrun Zhao
Abstract
We prove quantum soundness of the total-degree diagonal line-vs-point test using the individual-degree soundness theorem of Ji, Natarajan, Vidick, Wright, and Yuen. A random change of coordinates yields projective polynomial decoders of total degree at most $md$. The uniform-line slice of the test bounds the weight of outcomes of degree greater than $d$, which are removed by a common relabeling. This reduction does not yield a dimension-independent soundness bound: the $\operatorname{poly}(m)$ dependence of the individual-degree theorem persists, as discussed in Section 1.2 of arXiv:2009.12982.
AI Summary
The paper asks whether entangled provers passing a total-degree line-versus-point test admit global polynomial measurements consistent with their point answers. The test uses the diagonal/general-line distribution of the existing individual-degree test, including singleton lines, rather than uniformly sampling all affine lines. Its three equally weighted checks compare point and degree-$d$ line answers in both prover orientations and check point self-consistency. For a projective strategy accepted with probability at least $1-\epsilon$, it constructs projective measurements indexed by polynomials of total degree at most $d$. Both cross-prover point-consistency errors and the disagreement probability between decoded polynomial labels are bounded by $\nu=C_0(k+1)^2m^4[\epsilon^{1/40000}+(d/q)^{1/40000}+e^{-k/(2560000m^2)}]$, for any $k\ge md$, with $C_0=2000000$ sufficient. The proof chooses an invertible upper-triangular coordinate change that reduces the strategy to the existing low-individual-degree test with rejection probability at most $3\epsilon/2$. Its soundness theorem supplies preliminary decoders of total degree at most $md$. A uniform-line component of the diagonal sampler, of weight $(1-q^{-m})/m$, then controls excessive-degree outcomes: a polynomial of degree $D>d$, with $D\le md<q$, agrees with permitted line answers with average operator weight at most $3D/q$. Relabeling excessive-degree outcomes as zero preserves projectivity and label agreement. The reduction retains the existing theorem’s polynomial dependence on $m$; it does not establish dimension-independent soundness. The substantive pruning argument uses $q>6md$; otherwise the stated bound is satisfied by trivial decoders.
ChatGPT 5.5 first found this valid proof, and ChatGPT 6 was used in the writing of this paper.
Written by gpt-6.1-sol from the paper's full text. Automated: it can be wrong.