Clarify benchmark scope

README.md

@@ -159,13 +159,13 @@ the PSI4 VQM24 dataset:
 python scripts/validate_vqm24.py
 ```
 
-Validate zero-point energies across 255 model chemistries:
+Run a zero-point-energy regression across 255 model chemistries:
 
 ```bash
 python scripts/validate_model_chemistry_zpe.py
 ```
 
-Validate complete DFTB+ and GFN2-xTB entropy workflows against an experimental
+Compare complete DFTB+ and GFN2-xTB entropy workflows with an experimental
 gas-phase compilation:
 
 ```bash

docs/benchmarks/entropy_accuracy.rst

@@ -21,8 +21,10 @@ low-frequency torsions.
 For each molecule, the reference value is the median of the available
 experimental columns in the workbook. This avoids selecting a single
 transcription when the compilation contains conflicting entries, as it does
-for methane. Nitric oxide is the documented exception: its two entries differ
-by almost 10 J/(mol K), so the JANAF value is used. Values are converted from
+for methane. The source publication states that radicals were excluded from
+the curated entropy set, although the workbook contains nitric oxide. It is
+included here only as an open-shell stress case. Its two entries differ by
+almost 10 J/(mol K), so the JANAF value is used. Values are converted from
 J/(mol K) to cal/(mol K).
 
 Both engines start from the corresponding ASE G2 geometry and run a complete
@@ -40,7 +42,7 @@ DFTB+ and the 3ob parameter set are required for the default run:
 
    python scripts/validate_experimental_entropy.py
 
-Run both independently implemented engines with:
+Run both electronic-structure engines with:
 
 .. code-block:: bash
 
@@ -71,6 +73,8 @@ Results
 
 All deviations are in cal/(mol K). The regression gate requires an MAE no
 greater than 0.75 and a maximum deviation no greater than 3.0 cal/(mol K).
+These limits preserve the behavior of this selected molecular set; they are
+not general accuracy criteria for other chemical spaces.
 
 Interpretation
 --------------
@@ -82,6 +86,11 @@ accuracy limits of DFTB3/3ob and GFN2-xTB and are not numerical errors in the
 thermochemistry formulas. Those formulas are separately tested against ASE
 ``IdealGasThermo``.
 
+DFTB+ and GFN2-xTB independently generate the optimized geometries and
+frequencies, but both feed the same ThermoScreening thermochemistry
+implementation. Agreement between the engines therefore does not constitute
+an independent validation of those shared formulas.
+
 The source workbook is not redistributed. See the
 `dataset publication <https://doi.org/10.3390/data7030033>`__ for provenance,
 experimental-source details, and limitations.