Seungone Kim ✓

@seungonekim

Ph.D. student @LTIatCMU working on AI for science & LLM evals & nuclear fusion | Prev: @AIatMeta (FAIR) @kaist_ai @yonsei_u

1 961Followers

980Following

991Posts total

509.2KViews on collected posts

Growth & engagement

How the posts we collected actually performed: views and reaction rate post by post, what the audience did with them, and where the follower count goes.

Views per post

72.6K1 May

422.7K21 May

2.3K

4.1K

1.2K

2.1K

1.7K

895

802

889

Last 10 collected posts, oldest on the left. The scale is logarithmic: one post can outrun the rest a hundred times over.

Engagement rate per post

0.79%1 May

0.11%21 May

0.66%

0.29%

0.74%

0.73%

0.76%

1.45%

1.62%

1.24%

Reactions — likes, reposts, replies and quotes — divided by views.

What the audience does

Likes67.8%914 in total

Reposts13.2%178 in total

Replies2.1%28 in total

Quotes2.7%36 in total

Bookmarks14.3%193 in total

Share of every reaction we collected for this account. Replies mean argument, reposts mean endorsement, bookmarks mean the post was worth keeping.

The follower curve appears once this account has two daily snapshots — we take one a day, and this one is on its first.

Latest posts

@seungonekim very cool work @seungonekim! the AI reviewer overlap finding tracks what we found in our recent ICML 2026 paper: https://t.co/Fo8WN0R8mP 889 views · 8 likes · 2 reposts · 1 replies 21 May 2026 · Open on X →

Last but not least, I'd like to thank our co-authors for their fruitful feedback throughout this project. It took 9 full months until the release, I really appreciate all the help and advice throughout the journey ❤️ @dongkeun_yoon @kgashteo @scott_sjy @jinheonbaek 802 views · 13 likes · 0 reposts · 0 replies 21 May 2026 · Open on X →

More details are in the paper! https://t.co/OGArr8c7XP It also Includes 24 detailed case studies w/ full reviews and expert feedback on these reviews. https://t.co/zg8v2qaG2X 895 views · 10 likes · 2 reposts · 1 replies 21 May 2026 · Open on X →

We also compared all pairs of criticisms and found that two AI reviewers raise similar points to each other more often than two human reviewers do. Replacing a panel of human reviewers with AI would likely reduce the diversity of perspectives, a crucial characteristic of peer ht 1.7K views · 9 likes · 1 reposts · 2 replies 21 May 2026 · Open on X →

Do these results mean we can replace human reviewers with AI? Not yet! We got 1,017 qualitative comments from our expert annotators and surfaced 16 recurring weaknesses and 6 recurring strengths of current AI reviewers. The top-3 weaknesses are: 1. Missing community / field htt 2.1K views · 11 likes · 1 reposts · 2 replies 21 May 2026 · Open on X →

We also ran a pair-wise comparison: expert scientists were shown two reviews (one AI, one human), and asked which was better overall. GPT-5.2 is rated similar or better 48.6% and 73.4% of the time compared to the top/lowest rated human reviewer, respectively. https://t.co/oDYCgT 1.2K views · 7 likes · 1 reposts · 1 replies 21 May 2026 · Open on X →

On all three dimensions, AI reviewers raise more correct, significant, and well-evidenced criticisms than the lowest-rated human reviewer (the proverbial "Reviewer #2"). Against the top-rated reviewer, AI falls behind on correctness - but its correct criticisms tend to be more h 4.1K views · 8 likes · 1 reposts · 1 replies 21 May 2026 · Open on X →

So how do we got these results? Most prior work evaluates AI reviewers by a simple proxy: does the verdict (e.g., score or accept/reject) match a human's? 🤔But verdict matching doesn't mean the review is good. An AI can land at "5/10" while the specific criticisms it raises are 2.3K views · 11 likes · 3 reposts · 1 replies 21 May 2026 · Open on X →

Recently, there's been complaints on low-quality AI reviews at conferences and journals. What if we put the frontier LMs into an agent harness? With the right setup, on 82 Nature-family papers, 45 expert scientists judged that AI reviewers outperform the best human reviewer! 🤗 422.7K views · 367 likes · 91 reposts · 5 replies 21 May 2026 · Open on X →

Can you boost your AI review scores by asking an LLM to rewrite your paper? Yes! We call it paper laundering Our @icmlconf spotlight paper argues current AI reviewers aren't ready to automate peer review, and outlines what a science of peer review automation should look like🧵👇 ht 72.6K views · 470 likes · 76 reposts · 14 replies 01 May 2026 · Open on X →

Growth & engagement

Views per post

Engagement rate per post

What the audience does

Latest posts

Similar accounts