mlsanityData to 14 Oct 2026

Let Models Speak Ciphers: Multiagent Debate through Embeddings

arXiv 2310.06272 ↗First posted 10 Oct 2023cs.CL cs.AI cs.LGcited by 51

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

Abstract (arXiv)

Discussion and debate among Large Language Models (LLMs) have gained considerable attention due to their potential to enhance the reasoning ability of LLMs. Although natural language is an obvious choice for communication due to LLM's language understanding capability, the token sampling step needed when generating natural language poses a potential risk of information loss, as it uses only one token to represent the model's belief across the entire vocabulary. In this paper, we introduce a communication regime named CIPHER (Communicative Inter-Model Protocol Through Embedding Representation) to address this issue. Specifically, we remove the token sampling step from LLMs and let them communicate their beliefs across the vocabulary through the expectation of the raw transformer output embeddings. Remarkably, by deviating from natural language, CIPHER offers an advantage of encoding a broader spectrum of information without any modification to the model weights, outperforming the state-of-the-art LLM debate methods using natural language by 0.5-5.0% across five reasoning tasks and multiple open-source LLMs of varying sizes. This showcases the superiority and robustness of embeddings as an alternative "language" for communication among LLMs. We anticipate that CIPHER will inspire further exploration for the design of interactions within LLM agent systems, offering a new direction that could significantly influence future developments in the field.

Key facts. Cited by 51 papers in the map (data through 14 Oct 2026). Placed in Reasoning for agents, planning and code, one of 17 directions in LLM reasoning. Builds on 27 of its 44 references that are in the map. First posted on arXiv 10 Oct 2023.

Builds on

27 of 44 references.

Closest ancestors · 2

Same direction, newest first

  1. 2023Encouraging Divergent Thinking in Large Language Models through Multi-Agent Debate30 May
  2. 2023Teaching Large Language Models to Self-Debug11 Apr

Other references · 25

  1. 2023GPT-4 Technical Reportcited by 18,396
  2. 2023LLaMA: Open and Efficient Foundation Language Modelscited by 14,132
  3. 2023Llama 2: Open Foundation and Fine-Tuned Chat Modelscited by 13,028
  4. 2022Chain-of-Thought Prompting Elicits Reasoning in Large Language Modelscited by 12,459
  5. 2021Training Verifiers to Solve Math Word Problemscited by 10,109
  6. 2023Judging LLM-as-a-Judge with MT-Bench and Chatbot Arenacited by 8,626
  7. 2020Measuring Massive Multitask Language Understandingcited by 7,963
  8. 2022Self-Consistency Improves Chain of Thought Reasoning in Language Modelscited by 5,708
  9. 2022Large Language Models are Zero-Shot Reasonerscited by 4,374
  10. 2023Self-Refine: Iterative Refinement with Self-Feedbackcited by 3,442
  11. 2023Tree of Thoughts: Deliberate Problem Solving with Large Language Modelscited by 3,211
  12. 2022Constitutional AI: Harmlessness from AI Feedbackcited by 3,050
13 more
  1. 2023Improving Factuality and Reasoning in Language Models through Multiagent Debatecited by 1,784
  2. 2023WizardLM: Empowering large pre-trained language models to follow complex instructionscited by 1,125
  3. 2023Large Language Models are not Fair Evaluatorscited by 1,085
  4. 2023Graph of Thoughts: Solving Elaborate Problems with Large Language Modelscited by 886
  5. 2023The RefinedWeb Dataset for Falcon LLM: Outperforming Curated Corpora with Web Data, and Web Data Onlycited by 763
  6. 2023WizardMath: Empowering Mathematical Reasoning for Large Language Models via Reinforced Evol-Instructcited by 625
  7. 2022Self-critiquing models for assisting human evaluatorscited by 380
  8. 2023Large Language Model Guided Tree-of-Thoughtcited by 221
  9. 2023Improving Language Model Negotiation with Self-Play and In-Context Learning from AI Feedbackcited by 207
  10. 2023REFINER: Reasoning Feedback on Intermediate Representationscited by 205
  11. 2023What and How does In-Context Learning Learn? Bayesian Model Averaging, Parameterization, and Generalizationcited by 118
  12. 2023RL4F: Generating Natural Language Feedback with Reinforcement Learning for Repairing Model Outputscited by 105
  13. 2023SelfEvolve: A Code Evolution Framework via Large Language Modelscited by 43

Built on by

Cited by year (2026 to 14 Oct)

Citing papers per year: 2023: 1; 2024: 8; 2025: 18; 2026: 24.

51 papers.

Sort:

By area:

  1. 2024Training Large Language Models to Reason in a Continuous Latent Spacecited by 797
  2. 2024On scalable oversight with weak LLMs judging strong LLMscited by 99
  3. 2025Multiagent Finetuning: Self Improvement with Diverse Reasoning Chainscited by 95
  4. 2024From Individual to Society: A Survey on Social Simulation Driven by Large Language Model-based Agentscited by 82
  5. 2025Debate or Vote: Which Yields Better Decisions in Multi-Agent Large Language Models?cited by 81
  6. 2025Multi-Agent Verification: Scaling Test-Time Compute with Multiple Verifierscited by 62
  7. 2025Stop Overvaluing Multi-Agent Debate -- We Must Rethink Evaluation and Embrace Model Heterogeneitycited by 44
  8. 2025Enabling Agents to Communicate Entirely in Latent Spacecited by 34
  9. 2025Thought Communication in Multiagent Collaborationcited by 30
  10. 2025Think Smarter not Harder: Adaptive Reasoning with Inference Aware Optimizationcited by 29
  11. 2025From Debate to Equilibrium: Belief-Driven Multi-Agent LLM Reasoning via Bayesian Nash Equilibriumcited by 26
  12. 2024ACC-Collab: An Actor-Critic Approach to Multi-Agent LLM Collaborationcited by 24
38 more
  1. 2025Multi-Agent Debate for LLM Judges with Adaptive Stability Detectioncited by 23
  2. 2024Evaluating the Performance of Large Language Models via Debatescited by 19
  3. 2025KVComm: Enabling Efficient LLM Communication through Selective KV Sharingcited by 18
  4. 2025Literature Review Of Multi-Agent Debate For Problem-Solvingcited by 15
  5. 2024Beyond Natural Language: LLMs Leveraging Alternative Formats for Enhanced Reasoning and Communicationcited by 14
  6. 2024DebUnc: Improving Large Language Model Agent Communication With Uncertainty Metricscited by 14
  7. 2025Augmenting Multi-Agent Communication with State Delta Trajectorycited by 14
  8. 2025LLM as a Broken Telephone: Iterative Generation Distorts Informationcited by 12
  9. 2024Leveraging Large Language Models for Suicide Detection on Social Media with Limited Labelscited by 10
  10. 2025A Survey of Scaling in Large Language Model Reasoningcited by 10
  11. 2026The Vision Wormhole: Latent-Space Communication in Heterogeneous Multi-Agent Systemscited by 9
  12. 2025Multi-Agent Debate Strategies to Enhance Requirements Engineering with Large Language Modelscited by 8
  13. 2025CortexDebate: Debating Sparsely and Equally for Multi-Agent Debatecited by 7
  14. 2026Beyond tokens: a unified framework for latent communication in LLM-based multi-agent systemscited by 7
  15. 2023Playing Large Games with Oracles and AI Debatecited by 6
  16. 2026Do Latent Channels Actually Communicate? A Causal Audit of Latent Multi-Agent LLMcited by 5
  17. 2026StateBridge: Training-free Hidden-state Alignment for Latent Communication in LLM Multi-Agent Systemscited by 5
  18. 2025MV-Debate: Multi-view Agent Debate with Dynamic Reflection Gating for Multimodal Harmful Content Detection in Social Mediacited by 4
  19. 2026Hear Both Sides: Efficient Multi-Agent Debate via Diversity-Aware Message Retentioncited by 4
  20. 2026HyLaT: Efficient Multi-Agent Communication via Hybrid Latent-Text Protocolcited by 4
  21. 2026When Latent Agents Lie: KV-Cache Integrity in Multi-Agent LLM Collaborationcited by 4
  22. 2026KV Cache Translation across Heterogeneous Large Language Modelscited by 4
  23. 2026Yes FLoReNce, I Will Do Better Next Time! Agentic Feedback Reasoning for Humorous Meme Detectioncited by 3
  24. 2026Multi-Agent Debate Strategies: Survey, Taxonomy, and Challengescited by 3
  25. 2026XBridge: Entity-Grounded Latent Bridge for Heterogeneous LLM Communicationcited by 3
  26. 2025PD3: A Project Duplication Detection Framework via Adapted Multi-Agent Debatecited by 2
  27. 2026Cross-Modal Memory Compression for Efficient Multi-Agent Debatecited by 2
  28. 2026M3MAD-Bench: Multi-Dimensional Evaluation of Multi-Agent Debate Across Domains and Modalitiescited by 1
  29. 2026Confidence Laundering in Agent Systems: Why Uncertainty Needs a Latent Carriercited by 1
  30. 2026Translating the Translator: Decomposing the Cost of English-Forced Inter-Agent Communicationcited by 1
  31. 2026The impact of multi-agent debate protocols on debate quality: a controlled case studycited by 0
  32. 2026Good Agentic Friends Do Not Just Give Verbal Advice: They Can Update Your Weightscited by 0
  33. 2026The Periodic Table of LLM Reasoning: A Structured Survey of Reasoning Paradigms, Methods, and Failure Modescited by 0
  34. 2026Specification-Guided Synthesis of Deadlock-Free Communication Protocol Refinements with Large Language Modelscited by 0
  35. 2026AeroLat: Channel-Aware Latent Space Semantic Communication for Decentralized UAV Swarmscited by 0
  36. 2026Receiver-Conditioned Latent Communication gives 94% CacheBackcited by 0
  37. 2026Draft-KV: Learning Useful Latent Communication Between Language Modelscited by 0
  38. 2026Task-Oriented Key-Layer KV Communication for Efficient Latent Multi-Agent Collaborationcited by 0

50 most cited of 51 All 51 (JSON)

Similar, not cited

Close in wording, no citation either way

  1. 2023Preventing Language Models From Hiding Their Reasoning
  2. 2025Beyond Single Models: Enhancing LLM Detection of Ambiguity in Requests through Debate
  3. 2025Benchmarking Large Language Models for Cryptanalysis and Side-Channel Vulnerabilities
  4. 2025Security Concerns for Large Language Models: A Survey
  5. 2026When Embedding-Based Defenses Fail: Rethinking Safety in LLM-Based Multi-Agent Systems
  6. 2026The Implications of Linguistic Illegibility for LLM Security
  7. 2025S2-MAD: Breaking the Token Barrier to Enhance Multi-Agent Debate Efficiency
  8. 2026Think Before You Speak: Enabling Training-Free Latent Communication Between Language Models