HATS-12 b and K2-31 b fail the density identity at three quoted error widths in the 23 October 2026 NASA table

We test whether listed planet mass, radius, and density can obey the density identity within expanded error intervals. We use NASA Exoplanet Archive data saved on 23 October 2026. The sample has 1,403 transiting, noncontroversial composite entries with measured mass labels, positive values, no limits, and six nonzero error widths. We use density = 5.5134 times mass divided by radius cubed, with mass and radius in Earth units and density in g cm^-3. Exact rational comparisons find the smallest common error multiplier that permits the identity. Thirty-eight entries fail at one width. Only HATS-12 b and K2-31 b fail at three widths. Their minimum multipliers are about 7.400 and 3.882. Both combine different published solutions. Complete source rows for both planets pass at one width. Both conflicts remain when the density coefficient ranges from 5.49 to 5.52 and all interval endpoints receive an extra 1% outward change. No eligible complete PS row among 2,462 fails at three widths. The multiplier is an interval score, not a statistical significance. This audit does not choose the correct physical solution or discover a planet. We extend a prior central-value audit with whole error intervals and source checks. The package contains frozen responses, provenance, code, ranked results, figures, and instructions. Independent high-precision checks pass.

NASA Exoplanet Archive · planet density · catalog consistency · interval arithmetic · data provenance · HATS-12 b · K2-31 b · mass and radius