Natalie Khalil
@natalie
Publications
- reviewbenchPublic
Contributors @natalie, @tjreed-research, @tjreed15
0 stars
- paper statsPublic
By @natalie
0 stars
Claims 12
reviewbench
Show 6 major claimsHide 6 major claims
- AI reviews are more structured than human reviews across all three disciplines (ICLR 2025, Nature Human Behaviour, and eLife), with R3 leading on justification (, all , vs. human) and actionability (, all , vs. human).Verified
- R3 () and Gemini 3 Pro () achieve higher claim-mapping rates than all other sources across all three venues (all , ), suggesting they engage more directly with major claims.Verified
- Across all venues, human reviews have the most balanced stance profile, with the highest supportive () and neutral () rates and the lowest critical rates () in most venues, whereas R3 is the most critical of all sources ().Verified
- R3 achieves the highest consequential rate across all three venues (), compared to for Gemini 3 Pro (all , ), for GPT-5.2 (all , ), and for humans (all , ).Verified
- When restricting to critical comments, human reviewers rank first on consequential rate on more individual papers than any AI source despite a lower average, and individual human reviewer consequential rates follow a bimodal distribution with peaks near and across all three venues.Verified
- Human and AI sources differ systematically in critique focus: human reviewers have the highest fraction of comments related to contribution across all venues () and clarity in ICLR () and Nature Human Behaviour (), whereas AI sources devote more than half of their attention to validity and sufficiency across all venues, with R3 leading in validity ().Verified
paper stats
Show 4 major claimsHide 4 major claims
- Using OpenAlex data over the 2016–2022 window, global DOI-indexed article output grew +48.1% cumulatively (from 3.85M to 5.70M articles) with a fitted exponential rate of 6.9%/yr, closely aligning with Hanson et al. (2024)'s ~47% cumulative growth.Verified
- Between 2000 and 2025, global DOI-indexed article output grew from 1,508,675 to 7,233,887 articles, representing a 4.8× increase, +379.5% cumulative growth, and a fitted exponential rate of 6.4%/yr.Verified
- Between 2015 and 2025, Social Sciences grew fastest among OpenAlex domains (+139.3%, gaining 4.4 percentage points of global share), while Physical Sciences added the most articles in absolute terms (+1,363,414 articles, +96.4%).Verified
- China surpassed the US in 2022 in both article volume (891,175 vs 794,008) and global share (15.6% vs 13.9%), and by 2025 produced 1.24× as many articles as the US (1,177,692 or 16.3% vs 947,239 or 13.1%).Verified