The paper introduces a Mixture-of-Agents (MoA) approach to quantify the per-token compute required by large language models. By having a panel of fifteen models of varying sizes attempt to reproduce each token, the authors define the smallest successful agent’s inference cost as the token’s sufficient compute, providing an upper bound on necessary computation. Experiments on benchmarks show that a 0.5B model can reproduce most tokens, and that the MoA-derived compute map can reduce latency in model routing and drafting tasks while improving or maintaining accuracy.
By Zhixu Du, Weijia Han, Hai Helen Li, Yiran Chen
arXiv:2610.28619v1 Announce Type: new
Abstract: Multi-token prediction (MTP) can accelerate language-model decoding by drafting several tokens and verifying them in a single target-model pass. Howeve...
By Emile Anand, Mohammed Rehan Parwani, Xinyuan Cao
The paper presents a systematic study of token consumption in AI agents performing coding tasks. It finds that agentic tasks are far more expensive—about 1000 times more tokens than code reasoning or chat—primarily due to input tokens, and that token usage varies wildly, with accuracy peaking at moderate costs. Models differ significantly in efficiency, and current frontier models cannot reliably predict their own token usage, often underestimating it.
By Longju Bai, Zhemin Huang, Xingyao Wang, Jiao Sun, Rada Mihalcea, Erik Brynjolfsson, Alex Pentland, Jiaxin Pei
arXiv:2607. 14952v1 Announce Type: new Abstract: A growing gap separates inference context lengths from RL post-training: inference systems are approaching million-token contexts, while post-training workloads often remain at 256K tokens or below and rely on length generalization at deployment.
By Changhai Zhou, Kieran Liu, Yuhua Zhou, Qian Qiao, Jun Gao, Harry Zhang, Irvine Lu, Nolan Ho, Lucian Li, Andrew Lei, Cleon Cheng, Steven Chiang, Yihang Zeng, Di Zhang, Rio Yang, Kaijie Chen, Andrew Chen, Pony Ma, Weizhong Zhang, Cheng Jin
arXiv:2609.23790v1 Announce Type: new
Abstract: Every node in a multi-agent large language model (LLM) workflow retrieves context from memory and injects it into its prompt, where those injected toke...
By Vivek Kumar Singh, Preeti Priyam, Gautam Bhowmick
ForeSpark introduces a work‑conserving speculation (WCS) framework that jointly schedules draft and verification budgets for parallel speculative decoding in large language model serving. It uses a context‑aware pre‑draft estimator, a lightweight joint scheduler, and a post‑draft budget allocator to reduce wasted compute and improve throughput. Experiments on vLLM show up to 1.77× higher throughput than DSpark and 2.49× over autoregressive decoding, while maintaining the same output distribution and supporting higher request rates under latency SLOs.
arXiv:2608. 01428v1 Announce Type: cross Abstract: Embodied agents replan frequently to recover from execution drift, partial observability, and coordination hazards, but each LLM-based replanning call can consume an accumulated textual context that grows over time and across agents.
By Shuaijun Liu, Feiyang You, Xingwei Chen, Ningxin Su
arXiv:2610.02932v1 Announce Type: new
Abstract: Compiling GUI procedures that agents execute repeatedly into programs can reduce their token costs. However, measuring payback and deciding when to com...
By Yulong Ming, Jie Xu, Zihan Wu, Xiaohua Jia
arXiv:2608. 15592v1 Announce Type: new Abstract: Efficient LLM serving is often bottlenecked by the need to pad sequences to a fixed maximum length, and this wastes compute and degrades throughput.
By Feiyang Ren, Shengtao Wen, Lingbing Guo, Yu Tian, Yuanning Cui, Xiang Chen
The paper introduces CoPE, a method that trains language agents by adding multi‑step action targets to the policy backbone. By supervising each action forecasted from preceding histories, CoPE enables acting and forecasting to co‑evolve without extra inference cost. Across five benchmarks, CoPE outperforms existing RL baselines, achieving up to twice the sample efficiency on ScienceWorld and improving overall performance on realistic applications.
By Haofei Yu, Yixing Chen, Yining Zhao, Xuyan Ye, Yuxiang Liu, Zhigang Hua, Yan Xie, Shuang Yang, Jiaxuan You
The paper introduces Agentic Soft-Token Compression (ASTC), a training‑free method that uses soft tokens within an LLM agent to compress long‑horizon context. ASTC leverages the model’s recurrent dynamics and Gated Delta Network attention to identify absorbed information, forwarding it as soft tokens while keeping hard‑token values intact. Experiments show ASTC matches or surpasses uncompressed agents, reducing attention pairs by up to 50.68% and outperforming trained compressors by 15.4 to 103.7 times in attention usage.
By Xuning He, Kai Fan, Runyu Zhang, Binghuan Wang, Yuxiang Zhang, Xue Jiang, Yihong Dong
The paper introduces LRE (Learned Relevance Eviction), a lightweight, CPU‑only, language‑model‑free scorer that learns which parts of an agent’s interaction history are task‑critical and preserves them verbatim. In experiments, LRE matches or surpasses baseline eviction policies on accuracy‑cost trade‑offs, recovers 93% of full‑history accuracy, reduces worst‑case prompt size by 52%, and outperforms dense and token‑pruning encoders in conversational memory while being 295–1569× smaller. The method also achieves superior budgeted answer quality on LoCoMo reading and can be trained annotation‑free, recovering 95% of supervised scorer performance.
By Nusrat Jahan Lia, Aritra Mazumder