LLM-as-a-Verifier (arXiv

Key features

• Plus a user-adjustable verify timeout (default 90s — the ranking's own budget, never borrowed by sampling).
• Fine-grained rewardexpected grade over the top-20 logprob distribution, A=20…T=1 grouped band scale, grading at temperature 1.0 (the natural belief distribution — never collapsed)
• below 2 survivors the turn fails open to a normal answer.
• PPT pivot tournament (the paper's O(N·k) selection): random Hamiltonian ring (each candidate lands exactly once per slot — bias cancels inside the ring) → top-k pivots → only non-pivot×pivot pairs graded.
• Live-verifiedBo-5 grading calls 20 → 11 (−45%)