Models

← All models

Roth-Erev (Baseline) (Lech Mazur baseline strategy)

Lech Mazur Benchmarks

Roth-Erev is a non-LLM algorithmic baseline in Lech Mazur's BAZAAR double-auction benchmark. The official repository describes a reinforcement-learning strategy that maintains a propensity for each possible action and reinforces those propensities with rewards, making successful choices more likely; published game logs use `roth_erev` as its non-LLM `model_name`.

Benchmark scores