Multi-LLM Agent Collaborative Intelligence/ The Path to Artificial General Intelligence Edward Y. Chang

Format:

Book

Author/Creator:

Chang, Edward Y., author.

Series:

ACM books - Collection 3 ; #69.

ACM books, 2374-6777 ; #69

Language:

English

Subjects (All):

Multi-LLM Agent Collaborative Intelligence(Computer Science).

Genre:

Electronic books.

Physical Description:

1 PDF (xxvi, 570pages) LuaTEX

Edition:

First Edition

Place of Publication:

[New York, NY, USA] : Association for Computing Machinery; [2026].

System Details:

Mode of access: World Wide Web

System requirements: Adobe Acrobat Reader

Contents:

Preface

Acknowledgments

Acknowledgment of AI Assistance

Introduction to The MACI Framework: A System 2 Architecture for Artificial General Intelligence -The Intelligence Paradox

From Pattern Matching to Artificial General Intelligence

Eight Foundational Pillars

From System 1 Foundation to Cognitive AGI

Necessity and Sufficiency of the Eight Foundational Pillars

Book Organization

1 A Brief History of AI: From Turing to Transformers

Abstract

1.1 Introduction

1.2 Definitions

1.3 Perspectives on Similarity

1.4 Eras of Similarity Measurement

1.5 Concluding Remarks

References

2 Capabilities and Opportunities of Language Models

2.1 Introduction 2.2 Distinctive Capabilities

2.3 Exploring Unknown Unknowns

2.4 Conclusion

3 Prompt Engineering: Few Shots, Chain of Thought, and Retrieval-Augmented Generation

3.1 Introduction

3.2 Prompting Methods

3.3 RAG

3.4 Concluding Remarks

4 Unified Cognitive Consciousness Theory: Introduction

4.1 Introduction

4.2 Related Work

4.3 The UCCT Framework

4.4 Empirical Study: Qualitative Anchoring Demonstrations

4.5 Conclusion

5 CRIT: Socratic Inquiry for Critical Thinking in LLMs

5.1 Introduction

5.2 Related Work

5.3 The Socratic Method

5.4 Prompt Template Engineering

5.5 Pilot Study

5.6 Concluding Remarks

6 SocraSynth: Adversarial Multi-LLM Reasoning

6.1 Introduction

6.2 Multi-Agent SocraSynth Overview

6.3 Empirical Study

6.4 Remarks on Related Work

6.5 Concluding Remarks

6.6 Supplemental Materials

6.A Appendix A. Tables of Topic Proposal and Refinement

6.B Appendix B. Moderator's Prompts

6.C Appendix C. Debate from Opening to Fourth Round

6.D Appendix D. CRIT Evaluations by the Judges

6.E Appendix E. Debate Beats Q&A in Quality

6.F Appendix F. Post-Debate Conciliatory Remarks

6.G Appendix G. Healthcare Debate

7 EVINCE: Optimizing Adversarial LLM Dialogues via Conditional Statistics and Information Theory

7.1 Introduction

7.2 Related Work

7.3 The EVINCE Algorithm and Its Foundations

7.4 Empirical Evaluation of EVINCE

7.5 Concluding Remarks

7.A Appendix A: Metrics for LLM Debate Evaluation

7.B Appendix B: Theorem Proving

7.C Appendix C: Design Maxims of EVINCE Moderation

7.D Appendix D: EVINCE Debate #1

7.E Appendix E: EVINCE Debate-Jaundice vs. Hepatitis

8 Bias Correction and Data Synthesis through Reflective LLM Collaboration

8.1 Introduction

8.2 Related Work

8.3 Methodology

8.4 Experiments

8.5 Concluding Remarks

8.A Appendix A: On Annotation Quality

8.B Appendix B: Summary of EVINCE Debate on News D1

8.C Appendix C: Experiment #1 Justifications of Biased Articles

9 Modeling Emotions in Multimodal LLMs

9.1 Introduction

9.2 Qualifying and Quantifying Emotions

9.3 Empirical Study: Linguistic Features of Emotion

9.4 Qualifying and Quantifying Ethics

9.5 Concluding Remarks

9.A Appendix A

10 A Checks-and-Balances Framework for Context-Aware EthicalAI Alignment

10.1 Introduction

10.2 Related Work

10.3 Three-Branch Framework Design

10.4 Empirical Studies

10.5 Conclusion

10.6 Appendices

10.A Appendix A: Unconscious-Conscious Complementarity Thesis

10.B Appendix B: Wheels of Emotions

10.C Appendix C: Complex Emotions

10.D Appendix D: Hate Speech Samples

10.E Appendix E: Mixed Emotions

10.F Appendix F: Instruction to Human Annotators

10.G Appendix G: Polarized Emotions in One Article

10.H Appendix H: "To My Sister" of Different Linguistic Behaviors

11 SagaLLM: Persistent Context Management, Constraint Validation, and Transaction Guarantees

11.1 Introduction

11.2 Related Work

11.3 System Requirements for SagaLLM

11.4 Design with Travel Planning

11.5 Experiments

11.6 Conclusion

Supplementary Materials

Acknowledgment

12 ALAS: A Stateful Multi-LLM Agent Framework for Disruption-Aware Planning

12.1 Introduction

12.2 Related Work

12.3 The ALAS Three-Layer Architecture for Adaptive and Reactive Planning

12.4 Experimental Evaluation

12.5 Conclusion

12.6 Appendices

12.A Appendix A: Supplemental Information for Section 12.3

12.B Appendix B: Agent Factory Implementation Details

12.C Appendix C: LCPR Specification and Lemma Proofs

12.D Appendix D: Urban Ride Sharing

12.E Appendix E: Additional JSSP Results and Analysis

13 CoCoMo: Computational Consciousness Model

13.1 Introduction

13.2 Understanding Consciousness

13.3 Functionalities of Consciousness

13.4 Computational Consciousness

13.5 Concluding Remarks

Acknowledgments

14 Unified Cognitive Consciousness Theory: Formalization, Validation, and Outlook

14.1 Introduction

14.2 Background, Limitations, and Experimental Rationale

14.3 The UCCT Framework

14.4 Experiments: Validating UCCT

14.5 Conclusion

14.A Appendix A: Threshold-Crossing Dynamics Theorem

14.B Appendix B: Threshold Calibration

14.C Appendix C: Diversity Within Relevance

14.D Appendix D: UCCT for RAG (Protocol and Lab Exercise Plan)

14.E Appendix E: UCCT for Multi-Agent Debate (Protocol and Lab Exercise Plan)

15 A Retrospective and Adaptive Framework to Improve LLMs

15.1 Introduction

15.2 Related Work

15.3 Retrospective and Adaptive Learning

15.4 UCCT + RAFEL

15.5 Concluding Remarks

16 Polynthesis: Discovering Insights Beyond the Known

16.1 Introduction

16.2 Phase I: Warm-Up Breadth Probing

16.3 Phase II: From Breadth to Depth

16.4 Phase III: Concluding Remarks

16.5 Conclusion and Future Work

17 Aphorisms for Collaborative Intelligence

17.1 Introduction

17.2 List of 16 Aphorisms

17.3 Conclusion

Epilogue: Two Paradigm Bridges

Two Parallel Paradigm Shifts

Why the Pair Matters

Priority and Provenance

Acknowledgments

Author's Biography

Index

Other Format:

Print version:

ISBN:

3749421

9798400731723

9798400731730

Access Restriction:

Restricted for use by site license.