Extended Data Fig. 8: Shape-Only Reactor Spectral Distortion.

The data to prediction spectral ratio for the latest Double Chooz near detector (black), Daya Bay²⁹ (blue), RENO¹⁰ (red) and NEOS⁴⁵ (green) results are shown, exhibiting a common dominant pattern predominantly characterised by the 5 MeV excess. Small differences across experiments are still possible but unresolved so far. The Bugey3⁴¹ (not shown) is the only experiment known not to reproduce this structure. This remains an issue. The RENO and NEOS normalisation has been modified relative to publications to ensure the shape-only condition (average R = 1) is met. The 1σ uncertainty stands for 68% frequentist probability: both statistics (error bars) and the common reactor model prediction shape-only uncertainty (grey shaded). The shape-only is significantly smaller than the dominant rate-only uncertainties. Since the same reactor model prediction is used, this uncertainty is expected to remain a representative guideline to all experiments. The 5 MeV excess is compensated by a deficit region [1.5,4.0] MeV for all experiments due to the shape-only condition. A good agreement is found between Double Chooz and Daya Baya data throughout the entire energy range. The non-trivial match among different experiments suggests that most detector and part of the reactor effects are accurately reproduced by the simulation, thus cancelling across in R. This implies that the common reactor prediction model inaccuracies are expected to dominate the observed distortion. This is consistent with the fact that all other experiments use the same prediction strategy.