{"paper":{"title":"CHAL: Council of Hierarchical Agentic Language","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"CHAL reframes multi-agent LLM debate as structured belief optimization over domains where any position remains open to defeat by better reasoning.","cross_cats":["cs.LG","cs.MA"],"primary_cat":"cs.AI","authors_text":"Griffin D. Kent, Tommaso Giovannelli","submitted_at":"2026-05-12T20:26:41Z","abstract_excerpt":"Multi-agent debate has emerged as a promising approach for improving LLM reasoning on ground-truth tasks, yet current methodologies face certain structural limitations: debate tends to induce a martingale over belief trajectories, majority voting accounts for most observed gains, and LLMs exhibit confidence escalation rather than calibration across rounds. We argue that the genuine value of debate, and dialectic systems as a whole, lies not in ground-truth tasks but in defeasible domains, where every position can in principle be defeated by better reasoning. We present the Council of Hierarchi"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"CHAL is, to our knowledge, the first framework to treat multi-agent debate as structured belief optimization over defeasible domains.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"The assumption that LLMs can reliably maintain and perform gradient-informed revision on graph-structured CHAL Belief Schemas while treating meta-cognitive value systems as stable, effective hyperparameters without introducing inconsistencies or losing coherence.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"CHAL is a multi-agent dialectic system that performs structured belief optimization over defeasible domains using Bayesian-inspired graph representations and configurable meta-cognitive value system hyperparameters.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"CHAL reframes multi-agent LLM debate as structured belief optimization over domains where any position remains open to defeat by better reasoning.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"3669a0bf02e18d8348863e1fe659e11eaf278bbd942cf69626f7f26871edf495"},"source":{"id":"2605.12718","kind":"arxiv","version":1},"verdict":{"id":"2b01f054-4940-4c9c-ac02-38d8f896867a","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-14T19:54:03.141112Z","strongest_claim":"CHAL is, to our knowledge, the first framework to treat multi-agent debate as structured belief optimization over defeasible domains.","one_line_summary":"CHAL is a multi-agent dialectic system that performs structured belief optimization over defeasible domains using Bayesian-inspired graph representations and configurable meta-cognitive value system hyperparameters.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"The assumption that LLMs can reliably maintain and perform gradient-informed revision on graph-structured CHAL Belief Schemas while treating meta-cognitive value systems as stable, effective hyperparameters without introducing inconsistencies or losing coherence.","pith_extraction_headline":"CHAL reframes multi-agent LLM debate as structured belief optimization over domains where any position remains open to defeat by better reasoning."},"references":{"count":236,"sample":[{"doi":"","year":2025,"title":"E. Akata, L. Schulz, J. Coda-Forno, S. J. Oh, M. Bethge, and E. Schulz. Playing repeated games with large language models.Nature Human Behaviour, 9:1380–1390, 2025","work_id":"5cba845c-58fc-440b-80a3-6924468f3e42","ref_index":1,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":1985,"title":"C. E. Alchourrón, P. Gärdenfors, and D. Makinson. On the logic of theory change: Partial meet contraction and revision functions.The Journal of Symbolic Logic, 50:510–530, 1985","work_id":"260be2a8-a3b2-4de8-b80b-07f61d644cd5","ref_index":2,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2023,"title":"Claude: A large language model by anthropic","work_id":"3a1ffaf4-f3e2-48f8-8f98-5d1d7d06de63","ref_index":3,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2023,"title":"B. Arguello, E. S. Johnson, and J. L. Gearhart. A trilevel model for segmentation of the power transmission grid cyber network.IEEE Syst. J., 17:419–430, 2023","work_id":"99cca858-9467-4b42-9ed0-2b6767c808b9","ref_index":4,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":1985,"title":"Aristotle.Nicomachean Ethics. 1985. Translated by T. Irwin (Hackett Publishing, 1985)","work_id":"98520ddc-f488-4129-ab24-aa2116aec959","ref_index":5,"cited_arxiv_id":"","is_internal_anchor":false}],"resolved_work":236,"snapshot_sha256":"2f5c67a9a12335b0c386739e55ddda1ef4fdc64e0bcaaf915bf9ee981765b2c7","internal_anchors":17},"formal_canon":{"evidence_count":2,"snapshot_sha256":"97223cbe49e32f59abdbb276fbda2bdff74d9e4b1d5d4aa54629918bc40c868b"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"}