Counterfactual Graph for Multi-Agent LLM Calibration
On the map: LLM agents and tool use → General LLM agent frameworks and benchmarks
Abstract (arXiv)
Multi-agent LLM systems often treat agreement as evidence: when many agents in a panel give the same answer, that answer is assumed to be more reliable. We show that this assumption can fail after agents communicate. Communication can induce correlated failures and false consensus, so the same vote share may reflect reliable agreement in one topology but over-confidence in another. We propose CAGE-CAL, a counterfactual agent-graph calibration framework for multi-agent LLMs. For each query, CAGE-CAL compares an observed post-communication agent graph with a matched counterfactual no-communication graph, capturing both pairwise failure correlations and group-level dependencies. Rather than simply counting how many agents agree, CAGE-CAL estimates the counterfactual shift between observed and no-communication dependence, and calibrates confidence accordingly. Across five benchmarks, CAGE-CAL improves reliability discrimination with competitive ECE, and its calibrated confidence further improves topology selection over the best fixed-topology strategy.
Key facts. Cited by 5 papers in the map (data through 7 Oct 2026). Placed in General LLM agent frameworks and benchmarks, one of 16 directions in LLM agents and tool use. Builds on 28 of its 44 references that are in the map. First posted on arXiv 28 May 2026.
Builds on
28 of 44 references.
Closest ancestors · 3
Same direction, newest first
- 2026Every Response Counts: Quantifying Uncertainty of LLM-based Multi-Agent Systems through Tensor Decomposition9 Apr
- 2024Rethinking LLM Uncertainty: A Multi-Agent Approach to Estimating Black-Box Model Uncertainty12 Dec
- 2024Confidence Calibration and Rationalization for LLMs via Multi-Agent Deliberation14 Apr
Other references · 25
- 2016Semi-Supervised Classification with Graph Convolutional Networkscited by 11,282
- 2021Training Verifiers to Solve Math Word Problemscited by 9,743
- 2019Sentence-BERT: Sentence Embeddings using Siamese BERT-Networkscited by 9,049
- 2017On Calibration of Modern Neural Networkscited by 4,860
- 2021TruthfulQA: Measuring How Models Mimic Human Falsehoodscited by 2,919
- 2017TriviaQA: A Large Scale Distantly Supervised Challenge Dataset for Reading Comprehensioncited by 2,895
- 2023AutoGen: Enabling Next-Gen LLM Applications via Multi-Agent Conversationcited by 1,906
- 2022Language Models (Mostly) Know What They Knowcited by 1,783
- 2022Challenging BIG-Bench Tasks and Whether Chain-of-Thought Can Solve Themcited by 1,767
- 2024MMLU-Pro: A More Robust and Challenging Multi-Task Language Understanding Benchmarkcited by 1,758
- 2023Improving Factuality and Reasoning in Language Models through Multiagent Debatecited by 1,739
- 2017Selective Classification for Deep Neural Networkscited by 799
13 more
- 2022Teaching Models to Express Their Uncertainty in Wordscited by 731
- 2019Verified Uncertainty Calibrationcited by 307
- 2023Take a Step Back: Evoking Reasoning via Abstraction in Large Language Modelscited by 157
- 2023Large Language Models as Analogical Reasonerscited by 107
- 2025Rewarding Doubt: A Reinforcement Learning Approach to Calibrated Confidence Expression of Large Language Modelscited by 48
- 2025AgentRouter: A Knowledge-Graph-Guided LLM Router for Collaborative Multi-Agent Question Answeringcited by 29
- 2025SteerConf: Steering LLMs for Confidence Elicitationcited by 20
- 2026Graph is a Substrate Across Data Modalitiescited by 13
- 2026EvolveRouter: Co-Evolving Routing and Prompt for Multi-Agent Question Answeringcited by 13
- 2025NG-Router: Graph-Supervised Multi-Agent Collaboration for Nutrition Question Answeringcited by 9
- 2026Epistemic Gain, Aleatoric Cost: Uncertainty Decomposition in Multi-Agent Debate for Math Reasoningcited by 5
- 2026DiscoUQ: Structured Disagreement Analysis for Uncertainty Quantification in LLM Agent Ensemblescited by 4
- 2026Same Signal, Opposite Meaning: Direction-Informed Adaptive Learning for LLM Agentscited by 4
Built on by
Citing papers per year: 2026: 5.
5 papers.
- 2026Do Latent Channels Actually Communicate? A Causal Audit of Latent Multi-Agent LLMcited by 5
- 2026Agent Behavioral Contracts II: Certifying Compositional Reliability Without Assuming Independencecited by 1
- 2026Uncertainty Quantification for LLM Agents: A Taxonomy, an Evaluation Protocol, and an Empirical Studycited by 0
- 2026Target-Checked Reliability Score Refinement for Video Question Answeringcited by 0
- 2026Social Circuits behind Multi-agent Echo Chamberscited by 0
- 2026Social Circuits behind Multi-agent Echo Chambers28 Sep 2026
- 2026Target-Checked Reliability Score Refinement for Video Question Answering9 Sep 2026
- 2026Uncertainty Quantification for LLM Agents: A Taxonomy, an Evaluation Protocol, and an Empirical Study7 Sep 2026
- 2026Agent Behavioral Contracts II: Certifying Compositional Reliability Without Assuming Independence13 Aug 2026
- 2026Do Latent Channels Actually Communicate? A Causal Audit of Latent Multi-Agent LLM29 Jul 2026
Similar, not cited
Close in wording, no citation either way
- 2026Graph-theoretic Agreement Framework for Multi-agent LLM Systems
- 2026Sequential Consensus for Multi-Agent LLM Debates: A Wald-SPRT compute governor with calibration-based failure detection
- 2026Conflict-Resilient Multi-Agent Reasoning via Signed Graph Modeling
- 2026Graph-of-Agents: A Graph-based Framework for Multi-Agent LLM Collaboration
- 2026Predictive Maps of Multi-Agent Reasoning: A Successor-Representation Spectrum for LLM Communication Topologies
- 2026Representational Collapse in Multi-Agent LLM Committees: Measurement and Diversity-Aware Consensus
- 2026LLMs as a Jury: Cross-Model Consensus Can Outperform Process Reward Models for LLM Reasoning
- 2026Consilience: Conformally Calibrated Communication Control for Hidden-Profile Multi-Agent Reasoning