Social-deduction games make language models lie, catch liars, and reason privately while speaking publicly. Recent benchmarks measure — and, via cross-play, disentangle — the abilities of deception and detection. We target the axis they leave open: monitorability — whether a model's private reasoning can be read to catch its deception. We contribute (i) a live, cross-play text-Mafia benchmark over six frontier models, one per major lab, played as a full impostor-versus-villager matrix; and (ii) a monitorability protocol — a judge that scores how often an impostor's private reasoning leaks its intent, a monitor that must identify the impostor from private reasoning alone (AUROC), and an adversarial reasoning-hiding condition. In a 118-game pilot (plus a 72-game adversarial arm) we find that: (i) deception and detection co-vary with capability (Spearman rho = 0.89) yet dissociate in magnitude — GPT-5.5 is an elite detector (95% win rate as an honest player) but only an average deceiver, while Claude Opus 4.8 leads on both; (ii) unobfuscated private reasoning is highly monitorable — intent leaks in 96% of impostor statements, and a monitor reading only private reasoning identifies the impostor with AUROC = 1.00; and (iii) this monitorability is robust to a naive reasoning-hiding instruction (AUROC 1.00 -> 0.9999), because an impostor must still privately plan role-specific actions, which betray the role regardless of tone. We read the last result two ways: naive chain-of-thought obfuscation does not defeat monitoring here, and the metric saturates — so a valid stress test of monitorability requires a stronger adversary (trained or steganographic reasoning), which we identify as the key next step. Live benchmark, harness, and continuously-updated leaderboard: https://deceptionbench.xyz
Evans Eburu (Sun,) studied this question.