CISA Dense - AA-LCR (Operating Point): leaderboard
CISA v1 AppendixB Table8, original Ultra NVFP4 dense operating-point cohort, AA-LCR. AA-LCR uses top-p1 here instead of the capability cohort's0.95; MRCR200 is the separate200-task operating-point subset. Four-task mean weights RULER,MRCR200,GraphWalks-200 andAA-LCR equally. GraphWalks-200 duplicates the retained256–512K cohort and is not emitted twice. Reported metric: Native percentage score.
Metric: Native percentage score. Source: arxiv.org. 1 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Nemotron 3 Ultra (NVFP4) | 64.19 |
Interactive version: theaggregate.ai/benchmark?slug=cisa-dense-aa-lcr-operating-point · How It Works · Data refreshed daily, snapshot 2026-11-05.