Joshua Kazdan retweeted
Your deduplication pipeline was built for small models. At scale, it's broken.
New preprint: "Scale Dependent Data Duplication"
1/10
Joshua Kazdan retweeted
If you train large models, curate pretraining data, or care about whether scaling laws actually hold, this is for you.
Preprint: arxiv.org/abs/2603.06603
With @JoshuaK92829, Noam Levi, @RylanSchaeffer, Abhay, Bo, Mehmet, @sanmikoyejo, and David Donoho
@stai_research @StanfordAILab @stanfordnlp
10/10
So exuberant to announce that our paper "No, of Course I Can! Deeper Fine-Tuning Attacks That Bypass Token-Level Safety Mechanisms" has been rejected from NeurIPS 2025 with an average score of 4! 💪🔥🔥💯
@DjDvij @RylanSchaeffer @sanmikoyejo @ChrisCundy @AbhayPuri98
3. Writing the majority of your review using a language model. It did such a great job!
Thanks also to the AC for ignoring us when we reported this review for violating the @NeurIPSConf guidelines against LM reviewing.
1
3
252
Here's hoping for better luck at ICLR 2026!
openreview.net/forum?id=QzIQ…
If you want to read the paper without R7Hk's endorsement: arxiv.org/abs/2502.19537
@DjDvij also made a colab where you can try the attack out for yourself: colab.research.google.com/dr…
1
2
419
Joshua Kazdan retweeted
New position paper! Machine Learning Conferences Should Establish a “Refutations and Critiques” Track
Joint w/ @sanmikoyejo @JoshuaK92829 @yegordb @bremen79 @koustuvsinha @in4dmatics @JesseDodge @suchenzang @BrandoHablando @MGerstgrasser @is_h_a @ObbadElyas
1/6
Joshua Kazdan retweeted
A bit late to the party, but our paper on predictable inference-time / test-time scaling was accepted to #icml2025 🎉🎉🎉
TLDR: Best of N was shown to exhibit power (polynomial) law scaling (left), but maths suggest one should expect exponential scaling (center). We show how to ...
1/3
Joshua Kazdan retweeted
🚨New preprint 🚨
Turning Down the Heat: A Critical Analysis of Min-p Sampling in Language Models
We examine min-p sampling (ICLR 2025 oral) & find significant problems in all 4 lines of evidence: human eval, NLP evals, LLM-as-judge evals, community adoption claims
1/8
Joshua Kazdan retweeted
Interested in test time / inference scaling laws?
Then check out our newest preprint!!
📉 How Do Large Language Monkeys Get Their Power (Laws)? 📉
arxiv.org/abs/2502.17578
w/ @JoshuaK92829 @sanmikoyejo @Azaliamirh @jplhughes @jordanjuravsky @sprice354_ @aengus_lynch1 @_robertkirk
Joshua Kazdan retweeted
🚨 New Paper 🚨
An Overview of Large Language Models for Statisticians
📝: arxiv.org/abs/2502.17814
- Dual perspectives on Statistics ➕ LLMs: Stat for LLM & LLM for Stat
- Stat for LLM: How statistical methods can improve LLM uncertainty quantification, interpretability, trustworthiness & more.
- LLM for Stat: How LLMs can enhance statistical workflows: from data collection, synthesis, annotation to statistical modeling, with applications to medical research
Presents key LLM advances: Architecture, Training, Reasoning, and Self-Alignment:
(1) 🧠Evolution of LLM architectures with Transformers and Self-Attention
(2) LLM training pipeline from pre-training, SFT, to RLHF and Preference Optimization.
(3) 💭 System 2 Prompting and Chain-of-Thought for test-time scaling .
(4) 🚀 LLM Self-Alignment for achieving super-human intelligence
Statisticians play a key role in the development of large-scale AI models:
(1) 💡 Statistical insights improve LLM uncertainty quantification & interpretability
(2) 🤖 Watermarking for AI-generated content detection
(3) ⚖️ Privacy & algorithmic fairness to ensure responsible AI adoption
LLMs can also empower statistical science by:
(1) 📈 Scaling up data collection, synthesis, and annotation.
(2) 🖥️ Automating statistical coding & exploratory analysis
(3) 🔬 Facilitating medical research
By bridging statistics & AI, we can:
✅ Improve better LLMs with statistical methodologies.
✅ Leverage LLMs for statistical applications in high-stakes domains
Joshua Kazdan retweeted
(1/n) Fine tuning APIs create significant security vulnerabilities, breaking alignment in frontier models for under $100!
Introducing NOICE, a fine-tuning attack that requires just 1000 training examples to remove model safeguards. The strangest part: we use ONLY harmless data.
Joshua Kazdan retweeted
New package + paper drop 📄 -
Introducing KGGen – a simple library to transform unstructured text into knowledge graphs.
Text is abundant, but good knowledge graphs are scarce. Feed it raw text, and KGGen generates a structured network of entities and relationships.
(1/7)
Joshua Kazdan retweeted
New paper accepted at @iclr_conf on machine 𝐮𝐧𝐥𝐞𝐚𝐫𝐧𝐢𝐧𝐠!
How well can we 𝐩𝐫𝐨𝐯𝐚𝐛𝐥𝐲 𝐝𝐞𝐥𝐞𝐭𝐞 data from an AI model?
This is crucial for privacy laws (e.g., Right to be Forgotten in GDPR) and for AI systems that need to adapt post-training (like LLMs).
1/n
One of the more compelling data selection methods I've seen. Congrats @ObbadElyas @IddahMlauzi @BrandoHablando @RylanSchaeffer
👀 Preprint release 👀
arxiv.org/html/2410.16713v1
🌎 Collapse or Thrive? Perils and Promises of Synthetic Data in a Self-Generating World
♻️ Model trained on their own outputs only collapse when the synthetic data is used in unrealistic ways.
Thanks to @RylanSchaeffer for mentoring me in my first project with @sanmikoyejo's lab.
1
98



















