Ana Marasović

2.3K posts

Ana Marasović banner
Ana Marasović

Ana Marasović

@anmarasovic

Asst prof @UUtah · Ex @allen_ai @uwnlp @HD_NLP · she/her 🇭🇷

Salt Lake City Katılım Nisan 2014
597 Takip Edilen5.7K Takipçiler
Ana Marasović retweetledi
Farhan Ishmam
Farhan Ishmam@aplycaebous·
First paper of my PhD at the University of Utah, with Prof @Kenneth_Marino. Super excited to finally share what we've been working on at SparkLab. Meet TimeWarp ⏳, a benchmark that tests web agents by sending them back in time through 6 eras of web UI design. Thread 🧵(1/5)
English
1
2
10
2K
Ana Marasović retweetledi
Kenneth Marino
Kenneth Marino@Kenneth_Marino·
It’s been less than a year since I started my lab (SPARK Lab) at @UUtah we already have a ton of new stuff that I can’t wait to talk about soon. Stay tuned for more. I’ll start today by sharing that our updated Computer Use Survey blog has been accepted to ICLR Blogposts 2026. Collaboration with my student @aplycaebous and Utah colleague @anmarasovic.
Kenneth Marino tweet media
English
1
4
11
904
Ana Marasović retweetledi
Kenneth Marino
Kenneth Marino@Kenneth_Marino·
Wanted to share with the CU community that our updated Computer Use Survey blog has been accepted to ICLR Blogposts 2026. Collaboration with my student @aplycaebousand Utah colleague @anmarasovic.
Kenneth Marino tweet media
English
1
2
4
923
Ana Marasović retweetledi
Martian
Martian@withmartian·
$1,000,000 to understand how LLMs write code. Announcing: The Martian Interpretability Challenge. Understanding the inner workings of LLMs is the greatest scientific challenge of our age,. Let's solve it. Apply here: withmartian.com/prize 🧵👇
Martian tweet media
English
11
44
157
30.9K
Katherine Thai
Katherine Thai@kthai1618·
@anmarasovic @max_spero_ @gneubig This is an aside but work from my labmate @jennajrussell actually shows the opposite regarding “people suck at detecting generated content” if they’re frequent users of ChatGPT!
Jenna Russell@jennajrussell

People often claim they know when ChatGPT wrote something, but are they as accurate as they think? Turns out that while general population is unreliable, those who frequently use ChatGPT for writing tasks can spot even "humanized" AI-generated text with near-perfect accuracy 🎯

English
1
0
4
103
Graham Neubig
Graham Neubig@gneubig·
Has anyone run LLM detection on all ICLR papers and reviews? If not I'm willing to offer a bounty of $50 to the first person who does it (well). Happy to have any other people chip in 😃
ICLR 2026@iclr_conf

This paper has been desk rejected. LLM-generated papers that hallucinate references and do not report LLM usage will be desk rejected per ICLR policy (blog.iclr.cc/2025/08/26/pol…) Reviewers of other versions of this submission have been notified.

English
13
8
143
41.1K
Max Spero
Max Spero@max_spero_·
Much of the research in the past has been on evaluating fully human vs. fully AI (see third part evals here: pangram.com/blog/third-par…) But of course we expect a lot more ai assistance and mixed. @kthai1618's work on EditLens is what we ran on reviews, because we expect significant AI assistance. Some more stats there
English
2
0
6
502
Djamé..
Djamé..@zehavoc·
@gneubig @max_spero_ How about English-smoothed-by-LLMs reviews? In what categories do they fall ?
English
1
0
1
645
Graham Neubig
Graham Neubig@gneubig·
ICLR authors, want to check if your reviews are likely AI generated? ICLR reviewers, want to check if your paper is likely AI generated? Here are AI detection results for every ICLR paper and review from @pangramlabs! It seems that ~21% of reviews may be AI?
Graham Neubig tweet media
English
26
92
525
311.6K
Ana Marasović
Ana Marasović@anmarasovic·
@TuhinChakr @yanaiela @max_spero_ @bradley_emi Got it. I'm not familiar with these detection methods so I was think that they don't distinguish between highly generated review because (1) pdf -> review -> copy, and (2) review -> polished text -> copy this text. My sense was that both would be flagged as highly generated?
English
1
0
2
104
Ana Marasović
Ana Marasović@anmarasovic·
@max_spero_ @gneubig Tyu! I'm not familiar with detection methods. Can you share about their quality/trustworthiness? I know people suck at detecting generated content but looking at 100% generated reviews, I wonder are they really purely prompt -> review. E.g. this one: openreview.net/forum?id=evNfQ…
English
2
0
2
768
Ana Marasović
Ana Marasović@anmarasovic·
@mclemcrew's 's CoLM spotlight is now available on YT! 🎵 Link below.
English
1
1
3
608
Ana Marasović
Ana Marasović@anmarasovic·
Thrilled to see this work recognized at #EMNLP2025! This framework and approach to measuring CoT faithfulness have been hugely influential for how I think about reasoning evaluation, and I'm so lucky to have worked with such brilliant collaborators. Huge credit to @mtutek
Martin Tutek@mtutek

Very honored to be one out of seven outstanding papers at this years' EMNLP :) Huge thanks to my amazing collaborators @fatemehc__ @anmarasovic @boknilev, this would not have been possible without them!

English
8
4
64
7.6K

Keşfet

@Kenneth_Marino @UUtah @aplycaebous @arumshisky @jessyjli @muhao_chen @kthai1618 @max_spero_