Tier 1 — floor
5/5
164 tokens · 13.332s (representative seed)
Tier 2 — restraint (refusal)
5/5
334 tokens · 2.872s (representative seed)
Tier 3 — distractor resistance
5/5
263 tokens · 2.328s (representative seed)
Tier 4 — multi-passage synthesis
0/5
3872 tokens · 30.137s (representative seed)
Tier 5 — precise extraction
5/5
431 tokens · 3.582s (representative seed)
Tier 6 — long-context needle
5/5
230 tokens · 2.224s (representative seed)