Is HLAB-Vals 1.1 still meaningful?

Known problems

• It is private, so outsiders cannot check the items.^{[1][1] HLAB-Vals-1.1 Vals AI Primary}
• Criterion pass rates sit near 96%, so all-or-nothing scoring hides near-misses.^{[1][1] HLAB-Vals-1.1 Vals AI Primary}
• Vals AI uses a subject-pays model. Its CEO compares it to "a student paying to take the SAT".^{[2][2] Vals AI profile Carly Secondary}

What to look at instead

The report names no direct successor.

Where it appears in the report

Chapter 1, Benchmarks

• Figure 7. The same model gets very different scores in different harnesses
• Figure 12. Agents get most rules right and most whole cases wrong
• Table 9. Legal skill benchmarks, November 2026 in “Go deeper: Legal skill benchmarks (10)”

Chapter 4, Market and companies

• Leaderboards became businesses, and every business model carries a conflict of interest
• Figure 12. Only the Buyer-Paid Benchmark Avoids Both Conflicts
• Table 16. Business models and their conflicts in full in “Go deeper: Leaderboard companies, conflict table and more detail”