Recent advances in large language models (LLMs) have sparked growing interest in building fully autonomous agents. However, fully autonomous LLM-based agents still face significant challenges, including limited reliability due to hallucinations, difficulty in handling complex tasks, and substantial safety and ethical risks, all of which limit their feasibility and trustworthiness in real-world applications. To overcome these limitations, LLM-based human-agent systems (LLM-HAS) incorporate human-provided information, feedback, or control into the agent system to enhance system performance, reliability, and safety. These human-agent collaboration systems enable humans and LLM-based agents to collaborate effectively by leveraging their complementary strengths.This paper provides the first comprehensive and structured survey of LLM-HAS. It clarifies fundamental concepts, systematically presents core components shaping these systems, including environment and profiling, human feedback, interaction types, orchestration, and communication, explores emerging applications, and discusses unique challenges and opportunities arising from human-AI collaboration. By consolidating current knowledge and offering a structured overview, we aim to foster further research and innovation in this rapidly evolving interdisciplinary field. Paper lists and resources are available at https://github.com/HenryPengZou/Awesome-Human-Agent-Collaboration-Interaction-Systems.

Text-attributed graphs (TAGs) underpin many NLP applications, but real-world TAGs are often limited in scale, class balance, and diversity, motivating realistic and controllable TAG generation. Existing generators either replace node text with precomputed embeddings, reducing the task to vector-attributed graph generation, or rely on LLM prompting and agent simulation; neither learns a joint distribution over raw node text, edges, and labels. We propose a controllable substructure generator built from a lightweight causal LM. The model generates connected TAG substructures rather than whole graphs: each training substructure is serialized into lossless token sequences from multiple start nodes. Averaging next-token loss over these traversals treats traversal order as a latent variable and optimizes a variational lower bound on the permutation-invariant substructure marginal likelihood. Training combines continual pretraining for the unconditional distribution with control-prefix fine-tuning that conditions generation on target structural and label descriptors. At inference, a single autoregressive pass per prefix emits a parseable substructure, and the same generator supports distribution-aligned synthesis and boundary-aware augmentation through target-side adaptation alone. Across five real-world TAGs, it produces losslessly parseable graphs, matches or exceeds structure-only generators in structural fidelity, and improves cross-graph GNN generalization.