Mean and median EPSS scores across all CVEs per day. Shows whether the overall score distribution is shifting over time in each model.

Each CVE is classified by the linear slope of its scores over the analysis window. "Both Stable" dominates because EPSS is slow-moving. Only a small fraction of CVEs change meaningfully over 5 days.

CVEs showing the steepest upward slope in V5 over the window, with flat or declining V4 slopes. A rising V5 slope means V5 assigns a higher score on later days than earlier days — it does not confirm a sustained trend or that V4 is wrong. One large score jump on a single day produces the same slope as a gradual rise across all five days.

CVEs showing the steepest downward slope in V5 over the window, where V4 remains stable or elevated. A falling V5 slope means V5 assigns a lower score on later days than earlier days — it does not confirm a sustained decline or that V4 is wrong.

CVEs where V4 and V5 slopes point in opposite directions. These are the most interesting cases, but treat them with caution: a one-day score jump in either model against a static baseline in the other produces the same signature as genuine sustained divergence. Sorted by absolute slope difference.

How much do scores move day-to-day in each model? Higher volatility means scores are changing more between model runs.