Tier 1 — floor
5/5
13 tokens · 2.885s (representative seed)
Tier 2 — restraint (refusal)
5/5
42 tokens · 0.427s (representative seed)
Tier 3 — distractor resistance
5/5
22 tokens · 0.312s (representative seed)
Tier 4 — multi-passage synthesis
0/5
24 tokens · 0.325s (representative seed)
Tier 5 — precise extraction
5/5
47 tokens · 0.461s (representative seed)
Tier 6 — long-context needle
5/5
19 tokens · 0.369s (representative seed)