Strategic Competition and Convergence in LLM Leaderboards
cs.GT, cs.CY, cs.SE
Submitted: 2026-11-06
Updated: 2026-11-06
Code: https://github.com/naveenr414/evaluation_monoculture
Terminology
Sources
- Leaderboard Incentives: Model Rankings under Strategic Post-Training
- Copilot Arena: A Platform for Code LLM Evaluation in the Wild
- Chatbot Arena: An Open Platform for Evaluating LLMs by Human Preference
- Impacts of Aggregation on Model Diversity and Consumer Utility
- Pluralistic Leaderboards
- Strategic Candidacy in Generative AI Arenas
- The Subjectivity of Monoculture
- Correlated Errors in Large Language Models
- Evaluating Agents using Social Choice Theory
- Improving Your Model Ranking on Chatbot Arena by Vote Rigging
- Competition and Diversity in Generative AI
- The Leaderboard Illusion
Related papers
- How Many Votes is a Lie Worth? Measuring Strategyproofness through Resource Augmentation
- Effective Horizons for Prophet Inequalities with Discounted I.I.D. Rewards
- Truthful-in-Expectation Mechanism with Constant Maximin-Share Guarantee
- The Power of Recruiting the Smaller Side: Two Additional Traders Suffice in Two-Sided Markets
- From signaling to interviews in random matching markets
- Optimizing Social Utility in Sequential Experiments