
TT Faculty & Chief Scientist at CISPA | Director 💎 CRISTAL Lab (https://xitter.gq/t.co/N6HkzzXuAX) | Safe, Trustworthy & Human-Aligned AI | Prev @DSI_UChicago @UMichCSE
Chicago, IL
Joined July 2009
- Tweets389
- Following4.1K
- Followers1.1K
- Likes34.6K
🚀 Hiring PhDs, Postdocs & Interns in ML, LLM Alignment, Agentic AI & AI Safety at 💎 CRISTAL Lab @ CISPA!
Building foundational science behind safe, reliable & context-aware AI.
Apply / Details:
📌 career.cispa.de/jobs/cristal…
💎 cristal-lab.ai
Cristina Garbacea retweeted
🚀 Recruiting PhDs, Postdocs, and Interns for the 💎 CRISTAL Lab at CISPA!
If you are passionate about building safe and trustworthy AI, come join us! Applications are reviewed on a rolling basis with flexible start dates.
👉 Apply here: cristinagarbacea.com/cristal…
#AI #ML #Hiring
Cristina Garbacea retweeted
Verification is becoming a central bottleneck in AI. But a deeper problem arises (c.f. our work on replicating ICML papers).
How do we evaluate the verifiers?
Today, we're launching SAI Arena at @sailabshq, a platform for evaluating AI verification systems (broadly AI scientists) with feedback from the people who use them.
sai.science/arena
✨ Thrilled to announce that I am joining CISPA Helmholtz Center for Information Security as Tenure-Track Faculty & Chief Scientist starting September 1st!
Alongisde this, I am launching the CRISTAL Lab! 💎
🔗 cristinagarbacea.com
💎 At the CRISTAL Lab, our mission is to build human-aligned, safe, trustworthy, and continuously adaptable AI systems around key focus areas:
• Trustworthy AI • LLM Alignment • AI Robustness • Agentic AI & Continual Learning • Benchmarking & Evaluation
1
7
353
🚀 Recruiting PhDs, Postdocs, and Interns for the 💎 CRISTAL Lab at CISPA!
If you are passionate about building safe and trustworthy AI, come join us! Applications are reviewed on a rolling basis with flexible start dates.
👉 Apply here: cristinagarbacea.com/cristal…
#AI #ML #Hiring
Cristina Garbacea retweeted
How much of science is verifiable?
As AI agents become capable of increasingly complex, long-horizon tasks, we have an opportunity to rethink the research ecosystem. One possibility is to verify the science rather than only review the narrative in a paper, a clear blind spot in human peer review.
When ICML wrapped up, we used AI agents to review and replicate all 168 oral papers, including 105 full replications. We ask three questions:
• What does it cost to produce a top machine learning paper?
• What can replication reveal that narrative-only reviewing misses?
• How well does our review system align with existing human reviews?
What we found:
• The median estimated cost to reproduce every reported experiment was $8,900.
• Despite the cost, papers are rarely replicable. Only 7 reproduced more than 80%, and only 27 reproduced more than 40% of the claims we tested.
• Our system covered ~80% of the issues identified by two or more human reviewers.
Check out the full article. More links in the thread if you are interested!
#MachineLearning #AI #Science #PeerReview #Reproducibility #ICML
Current RLHF alignment forces LLMs to optimize for a hypothetical "average user" who represents no real person well.
In my new position paper accepted at #ICML2026, I argue why we must move from aggregated to personalized human preferences.
Moving to personalized LLMs isn't without risk -- it introduces challenges like filter bubbles and value lock-in. I propose a "bounded personalization" framework that preserves universal safety constraints while respecting individual variation.
1
3
277
Thrilled to present this at #ICML in July! It's an extra special milestone for me as it marks my first single-author paper.
Read the full preprint on arXiv here: arxiv.org/abs/2606.07629
#AIAlignment #LLMs #MachineLearning
3
254
Cristina Garbacea retweeted
Looking for a fun weekend read?
Introducing the Illustrated ICML 🌎
We indexed all 6000+ ICML papers and built a visual way to explore the whole landscape
Search your favorite topics, open a cluster, and dive right in
Inspired by @JayAlammar
When did you last check a leaderboard before picking an LLM? 🤔 Excited to share that our paper "Personalized Benchmarking: Evaluating LLMs by Individual Preferences" was accepted to ACL Findings 2026! 🎉 Joint work with Heran Wang and @ChenhaoTan
We found that for 57% of active Chatbot Arena users, individual rankings are statistically indistinguishable from a random ordering of models under Bradley-Terry. Users show substantial heterogeneity in topical interests and communication styles.
1
1
348
By modeling these features, we predict user-specific rankings and cut prediction error by up to 35% over aggregate baselines! 📉✨
Read the paper here: arxiv.org/abs/2604.18943
1
206
Cristina Garbacea retweeted
World models for better decisions.
Aug 31–Sep 2, hosted by @ChicagoBooth. WM@booth brings together CS, econ, finance + social science researchers, with keynotes from @ylecun and @Diyi_Yang.
CfP due June 12 → wm-booth.org
Cristina Garbacea retweeted
I find myself repeatedly explaining the difference between open-weight (DeepSeek), open-source (Olmo), open-development (Marin). Let's see if this restaurant analogy helps:
- Open-weight: food is made behind closed doors, server brings you the dish
- Open-source: food is made behind closed doors, server brings you the dish and the recipe
- Open-development: you see the chef make the dish in the kitchen (and can shout suggestions while its cooking)!
Cristina Garbacea retweeted
Diyi talking about what’s missing in human-AI collaboration, addressing the problems on both ends!
Excited to announce the 2026 iteration of the Communication & Intelligence Symposium at UChicago!
We have an amazing lineup of speakers @Diyi_Yang @johnhewtt @dashunwang @TomerUllman
We have a simple call for abstract that is due on Apr 15 (links 👇). Please come and share your research!
Co-organized with the awesome @universeinanegg and @divingwithorcas














