mlsanityData to 7 Oct 2026

When Upstream Messages Override Correct Answers: A Controlled Study of Multi-Agent LLM Collaboration

arXiv 2609.36855 ↗First posted 29 Sep 2026cs.AIcited by 0

On the map: LLM agents and tool use → Social simulation and multi-agent debate

Abstract (arXiv)

Multi-agent LLM systems rely on message passing among specialized agents to accomplish complex tasks. However, an upstream agent may provide useful information or an incorrect answer that causes a downstream agent to override a correct answer supported by its own evidence. Prior work has not clearly separated the benefits of communication from the damage caused by incorrect messages. We study this problem with controlled experiments across five benchmarks and five receivers, keeping the downstream task and evidence fixed while comparing answers under three conditions: no message, the upstream agent's original message, or a message with the opposite conclusion. Our experiments reveal three key findings. First, messages often help when the downstream agent would otherwise answer incorrectly. Second, messages can also hurt: when the downstream agent would answer correctly without a message, an incorrect upstream message changes the answer in up to 32% of cases. Third, in 94% of audited harmful cases, the downstream agent copies the upstream's specific wrong answer--a pattern we term answer substitution. Removing unreliable messages recovers part of the lost accuracy, suggesting that communication should be selective based on upstream reliability and the evidence already available to the downstream agent.

Key facts. No citing papers in the map yet (data through 7 Oct 2026). Placed in Social simulation and multi-agent debate, one of 16 directions in LLM agents and tool use. Builds on 39 of its 44 references that are in the map. First posted on arXiv 29 Sep 2026.

Builds on

39 of 44 references.

Closest ancestors · 4

Same direction, newest first

  1. 2026Misinformation Propagation in Benign Multi-Agent Systems15 Jun
  2. 2026The Bystander Effect in Multi-Agent Reasoning: Quantifying Cognitive Loafing in Collaborative Interactions11 May
  3. 2025Herd Behavior: Investigating Peer Influence in LLM-based Multi-Agent Systems27 May
  4. 2023Encouraging Divergent Thinking in Large Language Models through Multi-Agent Debate30 May

Other references · 35

  1. 2023Judging LLM-as-a-Judge with MT-Bench and Chatbot Arenacited by 8,393
  2. 2022ReAct: Synergizing Reasoning and Acting in Language Modelscited by 8,355
  3. 2023Toolformer: Language Models Can Teach Themselves to Use Toolscited by 4,072
  4. 2018HotpotQA: A Dataset for Diverse, Explainable Multi-hop Question Answeringcited by 3,686
  5. 2023Lost in the Middle: How Language Models Use Long Contextscited by 3,383
  6. 2023Self-Refine: Iterative Refinement with Self-Feedbackcited by 3,317
  7. 2023AutoGen: Enabling Next-Gen LLM Applications via Multi-Agent Conversationcited by 1,906
  8. 2023MetaGPT: Meta Programming for A Multi-Agent Collaborative Frameworkcited by 1,761
  9. 2023Improving Factuality and Reasoning in Language Models through Multiagent Debatecited by 1,739
  10. 2020Constructing A Multi-hop QA Dataset for Comprehensive Evaluation of Reasoning Stepscited by 1,259
  11. 2023Language Models Don't Always Say What They Think: Unfaithful Explanations in Chain-of-Thought Promptingcited by 1,081
  12. 2023Towards Understanding Sycophancy in Language Modelscited by 988
23 more
  1. 2023Large Language Models Cannot Self-Correct Reasoning Yetcited by 965
  2. 2021MuSiQue: Multihop Questions via Single-hop Question Compositioncited by 893
  3. 2019DROP: A Reading Comprehension Benchmark Requiring Discrete Reasoning Over Paragraphscited by 883
  4. 2023Large Language Models Can Be Easily Distracted by Irrelevant Contextcited by 820
  5. 2023ChatEval: Towards Better LLM-based Evaluators through Multi-Agent Debatecited by 724
  6. 2023Measuring Faithfulness in Chain-of-Thought Reasoningcited by 668
  7. 2023Can LLM Already Serve as A Database Interface? A BIg Bench for Large-Scale Database Grounded Text-to-SQLscited by 667
  8. 2023AgentVerse: Facilitating Multi-Agent Collaboration and Exploring Emergent Behaviorscited by 580
  9. 2025Why Do Multi-Agent LLM Systems Fail?cited by 559
  10. 2024AFlow: Automating Agentic Workflow Generationcited by 395
  11. 2025Reasoning Models Don't Always Say What They Thinkcited by 388
  12. 2024Automated Design of Agentic Systemscited by 360
  13. 2023Adaptive Chameleon or Stubborn Sloth: Revealing the Behavior of Large Language Models in Knowledge Conflictscited by 333
  14. 2023Chain-of-Verification Reduces Hallucination in Large Language Modelscited by 317
  15. 2023Simple synthetic data reduces sycophancy in large language modelscited by 174
  16. 2022Rich Knowledge Sources Bring Complex Knowledge Conflicts: Recalibrating Models to Reflect Conflicting Evidencecited by 142
  17. 2024Cut the Crap: An Economical Communication Pipeline for LLM-based Multi-Agent Systemscited by 129
  18. 2026Hallucination Cascade: Analyzing Error Propagation in Multi-Agent LLM Systemscited by 12
  19. 2024When Context Leads but Parametric Memory Follows in Large Language Modelscited by 10
  20. 2026Easier to Mislead Than to Correct: Harmful and Beneficial Revision in LLM Conformitycited by 6
  21. 2026The Hallucination Snowball: Modeling Error Propagation as State Transitions in Multi-Agent LLM Pipelinescited by 4
  22. 2026AnchorBench: A Multi-Pathway Benchmark for the Anchoring Effect in LLMscited by 2
  23. 2026Wrong but Useful: Trajectory Value Beyond Answer Correctness in Multi-Agent Messagescited by 1

Built on by

Not cited yet.

Similar, not cited

Close in wording, no citation either way

  1. 2026Beyond Final Accuracy: Auditing Communication in LLM Multi-Agent Systems
  2. 2026Emergent Misaligned Communication in Long-Horizon Multi-Agent LLM Commerce
  3. 2026Candidate supply and answer selection shape the value of LLM judging in multi-agent systems
  4. 2026Diagnosis Is Not Prescription: Linguistic Co-Adaptation Explains Patching Hazards in LLM Pipelines
  5. 2026Uncertainty Quantification for LLM Agents: A Taxonomy, an Evaluation Protocol, and an Empirical Study
  6. 2026One Axis, No Brake: Self-Knowledge Limits the Filtering of Harmful Peer Conformity in LLMs
  7. 2026Prospective Interpretation Risk: Principled Communication Control Between LLMs
  8. 2026Adversarial Feeds Steer LLM Agent Decisions Against Their Defaults