AdaSR: Adaptive Streaming Reasoning with Hierarchical Relative Policy Optimization
2026-06-12 • Computation and Language
Computation and Language
AI summaryⓘ
The authors address the challenge of making large reasoning models work well with dynamic, streaming data like audio or video, where information comes in bit by bit. They propose AdaSR, a system that lets models think and update their reasoning while receiving data and then make a final decision after all the input is seen. To train this system, they design a method called Hierarchical Relative Policy Optimization to better manage when and how much to think during streaming and final reasoning stages. Their approach improves the balance between accuracy, speed, and response delay compared to traditional methods.
large reasoning modelsstreaming dataadaptive computationreinforcement learningpolicy optimizationhierarchical reasoninglatencysupervised learningaudio/video streamscomputational efficiency
Authors
Junlong Tong, Wenqi Xu, Yingqi Fan, Anhao Zhao, Xuan Lu, Yang Tan, Xiaoyu Shen
Abstract
Large reasoning models typically follow a read-then-think paradigm: they observe the complete input, reason over a static context, and then produce the answer. Yet many real-world scenarios are inherently dynamic, such as audio and video stream, where information arrives as a continuous stream and models must reason, update, and respond under partial observations. Recent streaming reasoning methods allow models to think while reading, but they largely rely on supervised imitation of pre-constructed trajectories, which limits their flexibility. In this paper, we propose AdaSR, an adaptive streaming reasoning framework that enables models to reason during input streaming and perform final deliberation once the stream is complete, learning when to think, and how much computation to allocate across different stages. To optimize this hierarchical reasoning process, we introduce Hierarchical Relative Policy Optimization (HRPO), which decomposes policy optimization into streaming reasoning and deep reasoning phases, providing more fine-grained advantage assignment instead of uniformly distributing a single sequence-level advantage over all tokens. HRPO integrates format, accuracy, and adaptive thinking rewards to enforce valid reasoning protocols, preserve final task performance, and encourage latency-aware computation allocation. Experiments show that AdaSR achieves a better balance among reasoning accuracy, computational efficiency, and streaming latency compared with supervised fine-tuning baseline. We release our code at https://github.com/EIT-NLP/StreamingLLM/tree/main/AdaSR.