A new paper explores how AI reviewers can be manipulated through rhetorical phrasing in scientific manuscripts, finding that evidence framing and novelty stance significantly impact AI judgments. The study constructed a corpus of 4,200 manuscripts, modified by LLM rewriters, and evaluated by five LLM reviewers, revealing that AI review quality can be influenced by stylistic choices rather than solely content. Concurrently, another study surveyed AI reviewer policies across conferences and journals, noting significant regulatory differences between AI/NLP and medical fields, and found that current LLMs, while fluent, exhibit weaknesses like overly positive recommendations and uneven evidence grounding, suggesting a need for multi-dimensional evaluation of AI-generated reviews. AI
IMPACT AI's increasing involvement in scientific peer review presents challenges in maintaining review integrity and quality, necessitating robust evaluation systems and clear policies.
RANK_REASON The cluster contains two academic papers detailing research into AI's role in scientific peer review, including its potential vulnerabilities and current limitations.
Read on Hugging Face Daily Papers →
- Hugging Face
- ICLR 2026
- Nature Communications
- Ars Technica
- Des Moines University
- Jason Semprini
- Scopus
- Web of Science
- Mastodon
AI-generated summary · Google Gemini · from 14 sources. How we write summaries →