mlsanityData to 14 Oct 2026

Streaming Communication in Multi-Agent Reasoning

arXiv 2606.05158 ↗First posted 3 Jun 2026cs.CL cs.AI cs.MAcited by 0

On the map: LLM reasoning → Reasoning for agents, planning and code

also in 1 more
Abstract (arXiv)

Multi-agent reasoning systems adopt a "generate-then-transfer" paradigm that forces end-to-end latency to scale linearly with pipeline depth. We introduce StreamMA, a multi-agent reasoning system that streams each reasoning step to downstream agents as soon as it is generated, pipelining adjacent agents and thus reducing latency. Surprisingly, this pipelining also improves effectiveness: because multi-step reasoning quality is non-uniform and early steps are more reliable than later ones, working with these reliable early steps instead of the full chain prevents error-prone late steps from misleading downstream agents. We formalize both advantages with the first closed-form joint analysis of stream, serial, and single protocols, deriving the effectiveness ordering, speedup upper bound, and cost ratio. Across eight reasoning benchmarks spanning mathematics, science, and code, two frontier LLMs (Claude Opus 4.6 and GPT-5.4), and three topologies (Chain, Tree, Graph), StreamMA outperforms both baselines (avg. +7.3 pp, max +22.4 pp on HMMT 2026; Claude Opus 4.6-high). Beyond these contributions, we discover a "step-level scaling law": increasing per-agent steps consistently improves both effectiveness and efficiency, a new scaling dimension orthogonal to and composable with agent-count scaling.

Key facts. No citing papers in the map yet (data through 14 Oct 2026). Placed in Reasoning for agents, planning and code, one of 17 directions in LLM reasoning. Builds on 29 of its 32 references that are in the map. First posted on arXiv 3 Jun 2026.

Builds on

29 of 32 references.

Closest ancestors · 4

Same direction, newest first

  1. 2025Benefits and Limitations of Communication in Multi-Agent Reasoning14 Oct
  2. 2024Mixture-of-Agents Enhances Large Language Model Capabilities7 Jun
  3. 2024LiveCodeBench: Holistic and Contamination Free Evaluation of Large Language Models for Code12 Mar
  4. 2023Corex: Pushing the Boundaries of Complex Reasoning through Multi-Model Collaboration30 Sep

Other references · 25

  1. 2023Efficient Memory Management for Large Language Model Serving with PagedAttentioncited by 5,944
  2. 2023Let's Verify Step by Stepcited by 3,789
  3. 2023Self-Refine: Iterative Refinement with Self-Feedbackcited by 3,442
  4. 2023GPQA: A Graduate-Level Google-Proof Q&A Benchmarkcited by 2,879
  5. 2023Improving Factuality and Reasoning in Language Models through Multiagent Debatecited by 1,784
  6. 2022Fast Inference from Transformers via Speculative Decodingcited by 1,653
  7. 2023CAMEL: Communicative Agents for "Mind" Exploration of Large Language Model Societycited by 1,395
  8. 2023SGLang: Efficient Execution of Structured Language Model Programscited by 1,141
  9. 2023Accelerating Large Language Model Decoding with Speculative Samplingcited by 1,006
  10. 2023Math-Shepherd: Verify and Reinforce LLMs Step-by-step without Human Annotationscited by 928
  11. 2023ChatDev: Communicative Agents for Software Developmentcited by 919
  12. 2022Solving math word problems with process- and outcome-based feedbackcited by 901
13 more
  1. 2024Medusa: Simple LLM Inference Acceleration Framework with Multiple Decoding Headscited by 843
  2. 2021Efficient Large-Scale Language Model Training on GPU Clusters Using Megatron-LMcited by 632
  3. 2024Break the Sequential Dependency of LLM Inference Using Lookahead Decodingcited by 298
  4. 2023A Dynamic LLM-Powered Agent Network for Task-Oriented Agent Collaborationcited by 290
  5. 2025When More is Less: Understanding Chain-of-Thought Length in LLMscited by 183
  6. 2023Skeleton-of-Thought: Prompting LLMs for Efficient Parallel Generationcited by 105
  7. 2023Exchange-of-Thought: Enhancing Large Language Model Capabilities through Cross-Model Communicationcited by 82
  8. 2025Cache-to-Cache: Direct Semantic Communication Between Large Language Modelscited by 61
  9. 2025KVFlow: Efficient Prefix Caching for Accelerating LLM-Based Multi-Agent Workflowscited by 56
  10. 2025Group Think: Multiple Concurrent Reasoning Agents Collaborating at Token Level Granularitycited by 22
  11. 2026DualPath: Breaking the Storage Bandwidth Bottleneck in Agentic LLM Inferencecited by 15
  12. 2026ECHO: Elastic Speculative Decoding with Sparse Gating for High-Concurrency Scenarioscited by 12
  13. 2025Staircase Streaming for Low-Latency Multi-Agent Inferencecited by 3

Built on by

Not cited yet.

Similar, not cited

Close in wording, no citation either way

  1. 2025AgentsNet: Coordination and Collaborative Reasoning in Multi-Agent LLMs
  2. 2026DyTopo: Dynamic Topology Routing for Multi-Agent Reasoning via Semantic Matching
  3. 2026DIANOIA: Diagnostic Decomposition and Joint Optimization for Multi-Agent Reasoning
  4. 2026Second Thought: Reasoning in Parallel as LLM Agents Act and Observe
  5. 2026BusMA: A Bus Communication Substrate for Multi-Agent Systems
  6. 2026Multi-Agent Reasoning Improves Compute Efficiency: Pareto-Optimal Test-Time Scaling
  7. 2025Optimizing Sequential Multi-Step Tasks with Parallel LLM Agents
  8. 2026Learning to Communicate: Toward End-to-End Optimization of Multi-Agent Language Systems