X-ray bolometric correction, κ_bol, X = L_bol/L_{2 − 10 keV}, as a function of L_bol. Shown are LRDs (Yue et al. 2024, red circles), JWST-selected AGN (Maiolino et al. 2025, orange circles), BASS AGN (Gupta et al. 2024, blue crosses), SEAMBHs (Tortosa et al. 2023, yellow squares), and NLS1s (Jin et al. 2012b, cyan diamonds). The orange star marks the stacked X-ray measurement of the non-detected sources from Maiolino et al. (2025). Also shown are low z LRDs analogues (Lin et al. 2026, red empty circles), hyper-luminous QSOs at z ∼ 2 − 4 from the WISSH sample (Zappacosta et al. 2020; Degli Agosti et al. 2025, purple empty triangles), hyper-luminous QSOs at z > 6 from the HYPERION sample (black plus markers, Tortosa et al. 2024; Saccheo et al. 2025), SDSS DR7 QSOs at z ≃ 3.0 − 3.3 (Trefoloni et al. 2023, pink empty inverted triangle) and 0.4 < z < 3.3 radio-quiet highly accreting AGN (Laurenti et al. 2022, 2024, green diamonds). L_bol for BASS, SEAMBHs and NLS1 is from SED-fitting, while for LRDs and JWST-selected AGN L_bol is derived from the broad Hα luminosity following (Stern & Laor 2012), as in Maiolino et al. (2025). The black line indicates the relation from Duras et al. (2020), with the shaded region showing its 1σ uncertainty. The hatched region indicates the unphysical regime where κ_bol, X < 1.