STEM and Green Task Bundles
This page compares official occupation labels with postings-derived task-bundle evidence in the NLx corpus. STEM is the main case: the task signal is large enough to support a fuller comparison between official O*NET/BLS STEM occupations and postings-derived STEM candidates. Green is included below as a smaller companion check.
STEM And Green Task Intensity
LLM-validated task mentions are divided by NLx monthly active jobs. This normalizes by the MAJ denominator rather than raw posting volume.
The chart is included to show whether the two task-domain signals move at similar scale before the page turns to occupation classification.
STEM Task Bundle

Official STEM occupations
Posting-derived STEM candidates
Overlap
Official only
Posting-only
Corpus STEM task share
Table 1. Official STEM and postings-derived STEM candidates
STEM candidate screen
STEM_task_RCA = occupation STEM task share / corpus STEM task share
postings_derived_STEM_candidate = 1 when STEM_task_RCA > 1, active_jobs >= 25,000, STEM_task_mentions >= 250, distinct_STEM_task_IDs >= 5, STEM_mentions_per_active_job >= 0.08, and STEM_task_share >= 4%.
Table 2. Top occupations by LLM-validated STEM task RCA
Green Task Bundle Companion Check
Green task evidence moves in smaller steps than STEM in this NLx cut. The table below keeps the same official-vs-posting comparison, but the page treats it as a compact companion measure rather than a full standalone finding.

Official green occupations
Posting-derived green candidates
Overlap
Official only
Posting-only
Corpus green task share
Table 3. Top occupations by LLM-validated green task RCA
Evidence Path
STEM and green task labels use separate two-model agreement procedures documented in Agreement And Audit Trail. RCA is defined in Methods.