Tag
The paper proposes an unsupervised self-evolving agent framework inspired by diffusion models, using self-supervised semantic diffusion to train external skill libraries for LLM agents in specialized domains like creative screenwriting, without requiring weight access or external supervision.
This paper theoretically analyzes diffusion language models through a bias-variance lens, identifying trade-offs between masking and uniform diffusion kernels. It proposes SemDLM+, which adds a global transition and semantic-frequency penalty to overcome the semantic basin problem, achieving competitive generation quality on LM1B and OpenWebText benchmarks.