docs: cite the coupled-learning source protocol

2 files changed, 9 insertions, 0 deletions

diff --git a/README.md b/README.md
index 9cae3cb..9486ce2 100644
--- a/README.md
+++ b/README.md
@@ -35,6 +35,11 @@ three experimental parts are:
    nonideal local correlated-double-sampling, and a SPICE check of the local
    sample/hold/subtract primitive.
 
+The CLLN task, circuit equations, released task panel, and component-error
+scales follow [Dillavou et al. (2025)](https://arxiv.org/abs/2505.22887). The
+size ladder extends the same periodic circuit construction while keeping the
+released 4-by-4 tasks as its smallest endpoint.
+
 The current one-line method is
 
 ```text
diff --git a/THREE_PART_EVIDENCE.md b/THREE_PART_EVIDENCE.md
index eef1477..725753d 100644
--- a/THREE_PART_EVIDENCE.md
+++ b/THREE_PART_EVIDENCE.md
@@ -48,6 +48,10 @@ The periodic grid ladder uses side lengths `4, 8, 12, 16, 24, 32`, or 32 to
 initial gate field, and component draw. Size-dependent update exposure is
 selected from ideal coupled learning only.
 
+The circuit equations, released 4-by-4 task panel, and component-error scales
+come from [Dillavou et al. (2025)](https://arxiv.org/abs/2505.22887). Larger
+periodic grids preserve the same local device law and learning primitive.
+
 Confirmed final classification error:
 
 | Edges | Clean | Same-RMS noise | Raw | Static calibration | SDIL |