Large-scale A/B tests on social platforms suffer from inherent network interference, violating the Stable Unit Treatment Value Assumption (SUTVA) and distorting measured results. Industrial practices for mitigating network interference face a foundational trade-off. For scalability, they often rely on clustering static graphs, which serve as imperfect proxies for true interference pathways. Conversely, theoretically sound methods remain computationally intractable at production scale. This paper presents a production-ready framework deployed at TikTok, which integrates three core contributions to address these challenges: 1) Learned Interference Graph (LIG): Estimates interference probabilities using dynamic interaction patterns for more context-aware modeling. 2) Scalable Community Partitioning (SCP): A Spark-optimized ParLeiden implementation that performs billion-node graph clustering daily and generalizes effectively across diverse interaction types, achieving a purity score of 0.898 for group chat interactions. 3) Sensitivity-Enhanced Estimation (SEE): A multivariate system leveraging Controlled-experiment Using Pre-Experiment Data (CUPED) to mitigate variance inflation from cluster-based randomization. In live production tests, our framework reduces interference rates by 68.8%, correcting a biased treatment effect estimate from +1.44% to a statistically significant +2.08%. It also enables previously undetectable cross-ecosystem measurements, revealing a +0.2% lift in creator upload volumes driven by user-side treatments.

Live-streaming recommendation systems effectively match viewers' interests with relevant streamer content by ranking streamer candidates based on predicted relevance scores. However, existing works aggressively push popular streamers' content to match viewers' past interests, resulting in uneven traffic allocation. This creates a dual dilemma: viewers tend to concentrate on popular streamers, leading to reduced quality of experience due to streamers' limited capacity, while non-popular streamers suffer from traffic starvation. This paper formulates the viewer-streamer match problem as a capacity-aware optimal subgraph problem on the viewer-streamer bipartite graph, and proposes a novel capacity-aware ranking approach for live-streaming recommendation systems with explicit consideration of streamers' limited service capacity. Initially, we perform reinforcement learning to estimate the streamer capacity. Then, a novel graph algorithm FIMWS is introduced to solve the capacity-aware viewer-streamer match problem from a globally optimal perspective, with its theoretical optimality formally proven. The outcome of the proposed graph algorithm is finally distilled via deep models for point-wise online ranking. Currently, this capacity-aware ranking approach has been successfully deployed in TikTok Live. Extensive experiments on TikTok Live show that the proposed approach can significantly improve bilateral experience compared with other alternatives: +0.9% watch duration for viewers and +2% live duration for non-popular streamers.

Considering its advantages in dealing with high-dimensional visual input and learning control policies in discrete domain, Deep Q Network (DQN) could be an alternative method of traditional auto-focus means in the future. In this paper, based on Deep Reinforcement Learning, we propose an end-to-end approach that can learn auto-focus policies from visual input and finish at a clear spot automatically. We demonstrate that our method - discretizing the action space with coarse to fine steps and applying DQN is not only a solution to auto-focus but also a general approach towards vision-based control problems. Separate phases of training in virtual and real environments are applied to obtain an effective model. Virtual experiments, which are carried out after the virtual training phase, indicates that our method could achieve 100% accuracy on a certain view with different focus range. Further training on real robots could eliminate the deviation between the simulator and real scenario, leading to reliable performances in real applications.