Skip to main content
ModelScale

external benchmark

DiG-bench leaderboard

DiG-bench: Discovery in Games. Every model the catalog carries a published DiG-bench value for, ranked by that value.

Categoryexternal
MeasureExpected composite score with confidence interval
Tasks70 interactive games across seven difficulty tiers
DifficultyUnknown-rule discovery and interactive reasoning

Tests scientific discovery by asking agents to infer undisclosed rules across 70 interactive text-based games.

No model in the catalog has a published DiG-bench score.

The benchmark is defined by DiG-bench: Discovery in Games, but the catalog carries no value for it yet. An absent value is shown as absent here rather than as a zero.

All leaderboards