Deep20Bench B
AI benchmark · Game
- Tier
- B why
- Snapshot
What is Deep20Bench?
Deep20Bench evaluates how well language models play Twenty Questions. Each round gives a model a subject category and up to 50 yes-or-no questions to identify a hidden person or character, measuring question strategy, use of prior answers, and whether the model finishes before the turn limit. Scoring is lower-is-better: every question, wrong guess, and output-contract violation adds one point, and an unsolved round scores 51. Answers are adjudicated by an Oracle that searches the live web with citations, a blind Reviewer that decides each answer, and a blind Judge that resolves disagreements; the guesser sees only gameplay, never adjudicator prompts or evidence. The public pilot (protocol v9, cohort core-v2) covers 15 evaluated model configurations over 7 subjects with 5 rounds each, and also reports success rate, cost, runtime, and contract compliance.
Tier B
Established: 15 or more models scored, or curated RL directories rate it.
How much stands behind a tier: usage, who backs it, a confirmed line from its own site or paper, and a clear read on X. High with three of the four, medium with two, low otherwise.
- 18 models scored
- in curated RL directories (1 pt)
How we got to tier B
- Adoption: 18 models scored. Tier B.
- Not yet read on X.
- In curated RL directories: 1 point, a floor of none.
- Not yet confirmed by a line from its own site or paper.
A clear view on X moves a tier one step up or down; most entries don't move. How tiers work.
Not on the map yet: no line from its own site or paper places it in a domain.
Next: browse all benchmarks. Entry from the RL Research daily scrape of public sources, 2026-10-11 snapshot.