RESEARCHMonitorNEXT 12 MONTHS
SkillProx: Self-Evolving Agent Skills via Proximal Textual Gradient Descent
arXiv cs.CL — Computation and Language
Factual evidence
What the source reports
Researchers introduce SkillProx, an academic framework that dynamically refines and deletes agent skills in-context using textual gradient descent.
Open sourceOneBench interpretation
Institutional assessment
Hype caution
The paper assumes unconstrained runtime modification of agent prompts, which directly violates G-SIB model validation and change control frameworks.