Faster and Better LLMs via Latency-Aware Test-Time Scaling

• URL: http://arxiv.org/abs/2505.19634v4
• Date: Fri, 12 Sep 2025 01:41:20 GMT
• Title: Faster and Better LLMs via Latency-Aware Test-Time Scaling
• Authors: Zili Wang, Tianyu Zhang, Haoli Bai, Lu Hou, Xianzhi Yu, Wulong Liu, Shiming Xiang, Lei Zhu,
• Abstract summary: Test-Time Scaling (TTS) has proven effective in improving the performance of Large Language Models (LLMs) during inference.<n>Existing research has overlooked the efficiency of TTS from a latency-sensitive perspective.<n>We demonstrate that a compute-optimal TTS does not always result in the lowest latency in scenarios where latency is critical.
• Score: 47.3923926808606
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Test-Time Scaling (TTS) has proven effective in improving the performance of Large Language Models (LLMs) during inference. However, existing research has overlooked the efficiency of TTS from a latency-sensitive perspective. Through a latency-aware evaluation of representative TTS methods, we demonstrate that a compute-optimal TTS does not always result in the lowest latency in scenarios where latency is critical. To address this gap and achieve latency-optimal TTS, we propose two key approaches by optimizing the concurrency configurations: (1) branch-wise parallelism, which leverages multiple concurrent inference branches, and (2) sequence-wise parallelism, enabled by speculative decoding. By integrating these two approaches and allocating computational resources properly to each, our latency-optimal TTS enables a 32B model to reach 82.3% accuracy on MATH-500 within 1 minute and a smaller 3B model to achieve 72.4% within 10 seconds. Our work emphasizes the importance of latency-aware TTS and demonstrates its ability to deliver both speed and accuracy in latency-sensitive scenarios.

Related papers

• SWE-RM: Execution-free Feedback For Software Engineering Agents [61.86380395896069]
  Execution-based feedback is widely used in the development of coding agents through test-time scaling (TTS) and reinforcement learning (RL)<n>In contrast, execution-free feedback from reward models can provide more fine-grained signals without depending on unit test cases.<n>We introduce SWE-RM, an accurate and robust reward model adopting a mixture-of-experts architecture with 30B total parameters and 3B activated during inference.
  arXiv  Detail & Related papers  (2025-12-26T08:26:18Z)
• ThreadWeaver: Adaptive Threading for Efficient Parallel Reasoning in Language Models [99.6720868215076]
  We introduce ThreadWeaver, a framework for adaptive parallel reasoning.<n> ThreadWeaver achieves accuracy on par with popular sequential reasoning models of comparable size.<n>We show that ThreadWeaver delivers up to 1.53x average speedup in token latency.
  arXiv  Detail & Related papers  (2025-11-24T18:55:59Z)
• Nemotron-Flash: Towards Latency-Optimal Hybrid Small Language Models [97.55009021098554]
  This work aims to identify the key determinants of SLMs' real-device latency and offer generalizable principles and methodologies for SLM design and training.<n>We introduce a new family of hybrid SLMs, called Nemotron-Flash, which significantly advances the accuracy-efficiency frontier of state-of-the-art SLMs.
  arXiv  Detail & Related papers  (2025-11-24T08:46:36Z)
• AsyncSpade: Efficient Test-Time Scaling with Asynchronous Sparse Decoding [35.10915929939651]
  Test-time scaling (TTS) boosts LLM reasoning via long chain-of-thought (CoT)<n> KV-cache growth amplifies the memory-bound bottleneck of LLM decoding.<n>We propose AsyncSpade, an asynchronous framework for efficient TTS built on two core components.
  arXiv  Detail & Related papers  (2025-10-08T19:36:11Z)
• ATTS: Asynchronous Test-Time Scaling via Conformal Prediction [112.54016379556073]
  Large language models (LLMs) benefit from test-time scaling but are often hampered by high inference latency.<n>We introduce ATTS (Asynchronous Test-Time Scaling), a statistically guaranteed adaptive scaling framework.<n>We show that ATTS delivers up to 56.7x speedup in test-time scaling and a 4.14x throughput improvement.
  arXiv  Detail & Related papers  (2025-09-18T16:55:09Z)
• Slim-SC: Thought Pruning for Efficient Scaling with Self-Consistency [3.6199690908942546]
  Self-Consistency (SC) generates multiple reasoning chains in parallel and selects the final answer via majority voting.<n>We propose Slim-SC, a step-wise pruning strategy that identifies and removes redundant chains using inter-chain similarity at the thought level.<n> Experiments show that Slim-SC reduces latency and KVC usage by up to 45% and 26%, respectively, with R1-Distill.
  arXiv  Detail & Related papers  (2025-09-17T14:00:51Z)
• Towards Latency-Aware 3D Streaming Perception for Autonomous Driving [25.879279738510398]
  We propose a new benchmark tailored for online evaluation by considering runtime latency.<n>Based on the benchmark, we build a latency-aware 3D Streaming Perception framework.<n>Our method shows generalization across various latency levels, achieving an online performance that closely aligns with 80% of its offline evaluation.
  arXiv  Detail & Related papers  (2025-04-27T05:49:52Z)
• Learning Adaptive Parallel Reasoning with Language Models [70.1745752819628]
  We propose Adaptive Parallel Reasoning (APR), a novel reasoning framework that enables language models to orchestrate both serialized and parallel computations end-to-end.<n> APR generalizes existing reasoning methods by enabling adaptive multi-threaded inference using spawn() and join() operations.<n>A key innovation is our end-to-end reinforcement learning strategy, optimizing both parent and child inference threads to enhance task success rate without requiring predefined reasoning structures.
  arXiv  Detail & Related papers  (2025-04-21T22:29:02Z)
• Fast T2T: Optimization Consistency Speeds Up Diffusion-Based Training-to-Testing Solving for Combinatorial Optimization [83.65278205301576]
  We propose to learn direct mappings from different noise levels to the optimal solution for a given instance, facilitating high-quality generation with minimal shots.<n>This is achieved through an optimization consistency training protocol, which minimizes the difference among samples.<n>Experiments on two popular tasks, the Traveling Salesman Problem (TSP) and Maximal Independent Set (MIS), demonstrate the superiority of Fast T2T regarding both solution quality and efficiency.
  arXiv  Detail & Related papers  (2025-02-05T07:13:43Z)
• UniPTS: A Unified Framework for Proficient Post-Training Sparsity [67.16547529992928]
  Post-training Sparsity (PTS) is a newly emerged avenue that chases efficient network sparsity with limited data in need. In this paper, we attempt to reconcile this disparity by transposing three cardinal factors that profoundly alter the performance of conventional sparsity into the context of PTS. Our framework, termed UniPTS, is validated to be much superior to existing PTS methods across extensive benchmarks.
  arXiv  Detail & Related papers  (2024-05-29T06:53:18Z)
• Latency-aware Unified Dynamic Networks for Efficient Image Recognition [72.8951331472913]
  LAUDNet is a framework to bridge the theoretical and practical efficiency gap in dynamic networks. It integrates three primary dynamic paradigms-spatially adaptive computation, dynamic layer skipping, and dynamic channel skipping. It can notably reduce the latency of models like ResNet by over 50% on platforms such as V100,3090, and TX2 GPUs.
  arXiv  Detail & Related papers  (2023-08-30T10:57:41Z)
• Minimum Latency Training of Sequence Transducers for Streaming End-to-End Speech Recognition [38.28868751443619]
  We propose a new training method to explicitly model and reduce the latency of sequence transducer models. Experimental results show that the proposed minimum latency training reduces the latency of causal Conformer-T from 220 ms to 27 ms within a WER degradation of 0.7%.
  arXiv  Detail & Related papers  (2022-11-04T09:19:59Z)
• An Intelligent Deterministic Scheduling Method for Ultra-Low Latency Communication in Edge Enabled Industrial Internet of Things [19.277349546331557]
  Time Sensitive Network (TSN) is recently researched to realize low latency communication via deterministic scheduling. Non-collision theory based deterministic scheduling (NDS) method is proposed to achieve ultra-low latency communication for the time-sensitive flows. Experiment results demonstrate that NDS/DQS can well support deterministic ultra-low latency services and guarantee efficient bandwidth utilization.
  arXiv  Detail & Related papers  (2022-07-17T16:52:51Z)
• FastEmit: Low-latency Streaming ASR with Sequence-level Emission Regularization [78.46088089185156]
  Streaming automatic speech recognition (ASR) aims to emit each hypothesized word as quickly and accurately as possible. Existing approaches penalize emission delay by manipulating per-token or per-frame probability prediction in sequence transducer models. We propose a sequence-level emission regularization method, named FastEmit, that applies latency regularization directly on per-sequence probability in training transducer models.
  arXiv  Detail & Related papers  (2020-10-21T17:05:01Z)
• Good Feature Matching: Towards Accurate, Robust VO/VSLAM with Low Latency [23.443265839365054]
  Analysis of state-of-the-art VO/VSLAM system exposes a gap in balancing performance (accuracy & robustness) and efficiency (latency) This paper aims to fill the performance-efficiency gap with an enhancement applied to feature-based VSLAM.
  arXiv  Detail & Related papers  (2020-01-03T03:50:54Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.