Buildability / Receipt
This public receipt window renders only fields present in the canonical receipt object, deterministic fixture receipt, or canonical evidence receipt. Missing compute, demo, hash, signature, approval, telemetry, and adoption fields stay explicit.
Public buildability page receipt window
/buildability/benign-overfitting-in-adversarial-training-for-vision-transformers
Subject: Benign Overfitting in Adversarial Training for Vision Transformers
Verdict
Ignore
Verdict is Ignore because current viability and proof state do not clear the buildability gate.
Time to first demo
Insufficient data
No first-demo timestamp, owner estimate, or elapsed demo receipt is attached to this surface.
Data
C20 ϵ ≥ log ≥ log exp(V+(T2)) + C20 ϵ , |Vn,i(t)| = O(1). Thus, for n ∈ S+, we bound f( Xn,θ(t)) as follows : M φ( x⊤n,lWQWK⊤( Xn)⊤) XnWV wO ≥ log f( Xn
Truth Boundary
Buildability surfaces only report computed viability and proof receipts. They do not claim live production usage, pilot outcomes, founder sign-off, public Brier calibration, judge divergence, or external adoption unless explicitly sourced.
Compute
C20 ϵ ≥ log ≥ log exp(V+(T2)) + C20 ϵ , |Vn,i(t)| = O(1). Thus, for n ∈ S+, we bound f( Xn,θ(t)) as follows : M φ( x⊤n,lWQWK⊤( Xn)⊤) XnWV wO ≥ log f( Xn,θ) = M1 l=1 And ℓ(nt) = log 1 + exp(−f( Xn,θ(t))) ≤ exp(−f( Xn,θ(t))) ≤ exp −log 1 ϵ ≤ ϵ. Similarly, we have ℓ(nt) ≤ ϵ for n ∈ S−. Therefore
Inference
C20 ϵ ≥ log ≥ log exp(V+(T2)) + C20 ϵ , |Vn,i(t)| = O(1). Thus, for n ∈ S+, we bound f( Xn,θ(t)) as follows : M φ( x⊤n,lWQWK⊤( Xn)⊤) XnWV wO ≥ log f( Xn,θ) = M1 l=1 And ℓ(nt) = log 1 + exp(−f( Xn,θ(t))) ≤ exp(−f( Xn,θ(t))) ≤ exp −log 1 ϵ ≤ ϵ. Similarly, we have ℓ(nt) ≤ ϵ for n ∈ S−. Therefore
Hardware
C20 ϵ ≥ log ≥ log exp(V+(T2)) + C20 ϵ , |Vn,i(t)| = O(1). Thus, for n ∈ S+, we bound f( Xn,θ(t)) as follows : M φ( x⊤n,lWQWK⊤( Xn)⊤) XnWV wO ≥ log f( Xn,θ) = M1 l=1 And ℓ(nt) = log 1 + exp(−f( Xn,θ(t))) ≤ exp(−f( Xn,θ(t))) ≤ exp −log 1 ϵ ≤ ϵ. Similarly, we have ℓ(nt) ≤ ϵ for n ∈ S−. Therefore
Receipt path
/buildability/benign-overfitting-in-adversarial-training-for-vision-transformers
Paper ref
benign-overfitting-in-adversarial-training-for-vision-transformers
arXiv id
2604.19724
Generated at
2026-04-22T03:23:01.761Z
Evidence freshness
stale
Last verification
2026-04-22T03:23:01.761Z
Sources
3
References
0
Coverage
50%
Lineage hash
231c9010f4711f673fa2785fa51b7978de2f5c38f9d4bc9b0eff7030e068424f
Canonical opportunity-kernel lineage hash.
External signature
unsigned_external
No founder, registry, pilot, or production-adoption signature is attached to this receipt.
Verification
not_verified
Verification is blocked until an external signature is provided.
Some score or evidence fields are outside the preferred freshness window.
repo_url
references