The paper "The AI Review Lottery: Widespread AI-Assisted Peer Reviews Boost Paper Scores and Acceptance Rates" by Giuseppe Russo Latona, Manoel Horta Ribeiro, Tim R. Davidson, Veniamin Veselovsky, and Robert West examines the impact of AI-assisted peer reviews on the 2024 International Conference on Learning Representations (ICLR). The study addresses concerns about the potential negative influence of large language models (LLMs) on the validity and fairness of the peer-review system. Key findings include: 1. **Prevalence of AI-Assisted Reviews**: Using the GPTZero LLM detector, the study estimates that at least 15.8% of ICLR 2024 reviews were written with AI assistance. 2. **Impact on Paper Scores**: AI-assisted reviews were found to increase submission scores in 53.4% of pairs of reviews with different scores assigned to the same paper, with a relative difference of +14.4% in favor of AI-assisted reviews. 3. **Impact on Acceptance Rates**: AI-assisted reviews significantly boosted the acceptance rate of submissions, especially for those near the acceptance threshold. Submissions that received an AI-assisted review were 4.9 percentage points more likely to be accepted, corresponding to a 31.1% relative increase in the odds of acceptance. The study highlights the widespread use of AI-assisted reviews and their consequential impact on the peer-review process, suggesting that AI use can reduce the reliability and trust in the peer-review system. The authors call for ongoing measurements and guidelines to ensure the integrity and transparency of peer review in the evolving landscape of LLMs.The paper "The AI Review Lottery: Widespread AI-Assisted Peer Reviews Boost Paper Scores and Acceptance Rates" by Giuseppe Russo Latona, Manoel Horta Ribeiro, Tim R. Davidson, Veniamin Veselovsky, and Robert West examines the impact of AI-assisted peer reviews on the 2024 International Conference on Learning Representations (ICLR). The study addresses concerns about the potential negative influence of large language models (LLMs) on the validity and fairness of the peer-review system. Key findings include: 1. **Prevalence of AI-Assisted Reviews**: Using the GPTZero LLM detector, the study estimates that at least 15.8% of ICLR 2024 reviews were written with AI assistance. 2. **Impact on Paper Scores**: AI-assisted reviews were found to increase submission scores in 53.4% of pairs of reviews with different scores assigned to the same paper, with a relative difference of +14.4% in favor of AI-assisted reviews. 3. **Impact on Acceptance Rates**: AI-assisted reviews significantly boosted the acceptance rate of submissions, especially for those near the acceptance threshold. Submissions that received an AI-assisted review were 4.9 percentage points more likely to be accepted, corresponding to a 31.1% relative increase in the odds of acceptance. The study highlights the widespread use of AI-assisted reviews and their consequential impact on the peer-review process, suggesting that AI use can reduce the reliability and trust in the peer-review system. The authors call for ongoing measurements and guidelines to ensure the integrity and transparency of peer review in the evolving landscape of LLMs.