AI Briefing#robustness
The latest AI and developer news about #robustness, with the original source and a short summary.
The latest AI and developer news about #robustness, with the original source and a short summary.
This paper analyzes the limitations of existing machine learning from a Causal Inference perspective and presents the need for causal representation learning.
Using GPT-Red, an automated security tester developed by OpenAI, vulnerabilities in AI models were found and the safety of GPT-5.6 was improved by 6x.
RL-finetuned VLMs show improved visual reasoning performance, but exhibit vulnerability to textual perturbations.
An analysis of Meta's V-JEPA 2.1 models confirmed a non-monotonic scaling property in which robustness does not necessarily improve as model size increases.
A theoretical study reveals a structural blind spot in standard ERM and its worsening under PGD.