analiza aiqbench
This arXiv preprint proposes LIT, a multi-agent framework in which a video director agent, a simulated user, and a meta engineer co-evolve to improve both production quality and user interaction. The reported gains (19.1% on production and 32.0% on interaction scores over the initial pipeline) come from internal test splits, so they show relative improvement within this setup rather than performance against established baselines or real users. For AI readers, the notable idea is the closed loop in which the meta agent proactively acquires external knowledge instead of passively attributing failures.
Learning to Interact and Direct: Co-Evolving Video Director Agents
arXiv:2610.41865v1 Announce Type: new Abstract: Video director agents orchestrate multistage production to realize users' creative visions. Emerging research on their self-evolution advances screenplay planning effectively. However, the mutual reinforcement between directing and user interaction remains largely overlooked. Improved interaction clarifies user intent and informs workflow refinement, while more reliable execution shifts collaboration from troubleshooting to creative decisions. To this end, we introduce LIT (Learning to Interact and DirecT), a framework comprises a director agent, a simulated user, and a meta engineer. Simulated user assigns creative tasks to probe the director's weaknesses, while the meta engineer uses the task traces to revise the director's interaction and production strategy. Although this approach can address discovered interaction and production issues, it still relies on passive attribution by the meta agent. This limits its ability to adapt to broader user interaction patterns and production techniques. To solve this problem, we equip LIT with a co-evolution paradigm. The meta agent proactively learns external knowledge, thereby improving its ability to discover problems and propose solutions. The meta agent, simulated user, and director together form an evolutionary closed loop, collaboratively enhances the system's directing and interaction capabilities. Across the Test and OOD Test splits, LIT improves production and Interaction scores by 19.1% and 32.0%, respectively, relative to the initial pipeline. Controlled ablations reveal the correlation between improvements in production and interaction, where the co-evolution mechanism further encourage directing skill and interaction mode exploration.
Abstract:Video director agents orchestrate multistage production to realize users' creative visions. Emerging research on their self-evolution advances screenplay planning effectively. However, the mutual reinforcement between directing and user interaction remains largely overlooked. Improved interaction clarifies user intent and informs workflow refinement, while more reliable execution shifts collaboration from troubleshooting to creative decisions. To this end, we introduce LIT (Learning to Interact and DirecT), a framework comprises a director agent, a simulated user, and a meta engineer. Simulated user assigns creative tasks to probe the director's weaknesses, while the meta engineer uses the task traces to revise the director's interaction and production strategy. Although this approach can address discovered interaction and production issues, it still relies on passive attribution by the meta agent. This limits its ability to adapt to broader user interaction patterns and production techniques. To solve this problem, we equip LIT with a co-evolution paradigm. The meta agent proactively learns external knowledge, thereby improving its ability to discover problems and propose solutions. The meta agent, simulated user, and director together form an evolutionary closed loop, collaboratively enhances the system's directing and interaction capabilities. Across the Test and OOD Test splits, LIT improves production and Interaction scores by 19.1% and 32.0%, respectively, relative to the initial pipeline. Controlled ablations reveal the correlation between improvements in production and interaction, where the co-evolution mechanism further encourage directing skill and interaction mode exploration.
Artificial Intelligence (cs.AI)
arXiv:2610.41865 [cs.AI]
(or arXiv:2610.41865v1 [cs.AI] for this version)
https://doi.org/10.48550/arXiv.2610.41865 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Text integral sub licența deschisă a publicației · 325 cuvinte
Subiecte conexe
What Should a Large Language Model See? Physical Invariants as a Data Representation for PDE Discovery
arXiv · 27 aug. 2026
Self-Modifying Lean Proof Agents with Verifier-Grounded Benchmark Coevolution
arXiv · 21 iul. 2026
Multi-Objective Compliance-Integrated Coevolution For Simulated And Real-World Deployment Of Multi-Robot Marine Autonomy
arXiv · 30 iul. 2026
Task-CoEvolve: Efficient Harness Optimization via Adaptive Validation Task Selection
arXiv · 21 aug. 2026