In an experiment with synthetic jailbreaks, Constitutional Classifiers dramatically reduced jailbreak effectiveness. They increased refusal rates by a small amount (+0.4%) and increased compute overhead (+24%). We're working on reducing these costs.
RESEARCH
-

Claude Develops Constitutional Classifiers Against LLM Jailbreaks
By
–
Like all LLMs, Claude is vulnerable to jailbreaks—inputs designed to bypass its safety training and force it to produce outputs that might be harmful. Our new technique is a step towards robust jailbreak defenses. Read the blog post: https://
anthropic.com/research/const
itutional-classifiers
… -
Anthropic releases constitutional classifiers against universal jailbreaks
By
–
New Anthropic research: Constitutional Classifiers to defend against universal jailbreaks. We’re releasing a paper along with a demo where we challenge you to jailbreak the system.
-
AI Scaling Laws Remain Valid OpenAI DeepSeek
By
–
I don't think AI scaling laws have been repealed. @OpenAI @deepseek_ai
-
AI Models Face Challenges in Real-World Medical Conversations
By
–
AI models struggle in real-world medical conversations
#AI #AIio #AIInnovation #ML #DataScience #Futureofwork @Scobleizer @AndrewYNg @drfeifei @KirkDBorne @fchollet @rowancheung @antgrasso -
Open Research and Open Source Drive Tech Innovation Forward
By
–
If open research and open source didn't exist, almost *none* of the innovations we've seen over the last dozen years would have appeared.
No one would be ahead.
Everyone would be held back. -

Microsoft FP4 Quantization Framework Achieves Comparable Accuracy to FP8
By
–
5/ Microsoft introduces their framework for FP4 quantization training. FP4 was regarded unfeasible due to too much degradation, but the proposed method in this paper achieves accuracy that is comparable to FP8 and BF16. This means less computational demand to use LLMs, while
-

Top AI Papers: RL, SFT, Hallucination, FP4 Quantization
By
–
Top AI Papers you Can't miss from last Week: We summarized everything from RL vs. SFT training, Hallucination Mitigation with Multi-agent systems, Microsoft's New FP4 Quantization, and more. Here's everything you need to know:
-
Agentic Performance and GAIA Benchmark Analysis
By
–
With an agentic setup, quite obvious that scores would increase! The comparison against other agents on the GAIA benchmark is more interesting.
-
Appending Wait to Model Generation as Path to AGI
By
–
« appending "Wait" multiple times to the model's generation » => current most likely path to AGI
