PingPong: A Benchmark for Role-Playing Language Models with User Emulation and Multi-Model Evaluation
This provides a foundation for robust evaluation of language models in interactive scenarios, though it is incremental as it builds on existing benchmarking approaches.
The authors tackled the problem of evaluating role-playing capabilities in language models by creating a benchmark called PingPong, which uses simulated users and multi-turn conversations to assess over 40 models in English and Russian, showing strong correlation between automated and human evaluations.
We introduce a benchmark for evaluating the role-playing capabilities of language models. Our approach leverages different language models to simulate users in dynamic, multi-turn conversations and assess the resulting dialogues. Our methodology involves three main components: a player model that adopts a specific character role, an interrogator model that simulates user behavior in a specific situation, and a judge model ensemble that evaluates conversation quality with 3 metrics: character consistency, entertainment value, and language fluency. We evaluated more than 40 models in both English and Russian, with each model participating in 64 conversations with 8 characters and 8 situations. We conducted experiments comparing automated evaluations with human annotations to validate our approach, demonstrating strong correlations across multiple criteria. This work provides a foundation for a robust and dynamic evaluation of different model capabilities in interactive scenarios.