Hierarchical Certified Semantic Commitment for Byzantine-Resilient LLM-Agent Collaboration

2026-06-05

AI Summary

This work addresses the lack of fine-grained finality mechanisms for semantically rich, structured natural language proposals in Byzantine collaborative settings among existing large language model agents. The paper proposes the Hierarchical Certified Semantic Commitment protocol (H-CSC), which introduces, for the first time, a typed finality mechanism that outputs one of three outcomes—semantic commit, adjudicated commit, or abort with reason—based on semantic clustering and voting consistency. H-CSC integrates embedding-space clustering, 2f+1 semantic core detection, quantized summary generation, and Byzantine fault-tolerant logic, thereby overcoming traditional BFT’s reliance on byte-level agreement. Experiments demonstrate that H-CSC achieves reliable commits with low angular deviation (0.31–2.04°) and 100% aborts in infeasible rounds on BCS_v1; on the MVR-50 benchmark, it maintains high commit rates (0.90/0.92) under Byzantine attacks, exhibits negligible inefficiency (≤0.02), and generates verifiable semantic summaries in 74%/72% of rounds.

Abstract

Byzantine collaboration among large-language-model agents requires a finality-control primitive: given delivered stochastic, structured natural-language proposals, the protocol must decide whether the round supports a commit, what kind of commit, or a typed safe abort. Naive aggregation hides this choice behind a single verdict; classical Byzantine fault tolerance hides it behind byte-identity that LLM proposals do not satisfy. We introduce Hierarchical Certified Semantic Commitment (H-CSC), a BFT-inspired protocol that converts embedding-derived finality signals over verdict-conditioned proposal groups into one of three typed outcomes: a semantic_commit (a 2f+1 within-verdict semantic core backs the verdict, emitting a parameter-bound digest over the quantised aggregate), a verdict_commit (strong verdict margin but dispersed semantic rationale, emitting a verdict-level certificate without claiming a semantic aggregate), or an explicit abort with a typed reason. The contribution is typed finality, not raw commit accuracy. On a controlled semantic-poisoning diagnostic (BCS_v1, 120 episodes), H-CSC commits with low angular deviation on BFT-feasible buckets (0.31 to 2.04 degrees) and aborts 100% of beyond-BFT rounds (n<3f+1) as intended. On a real LLM-agent claim-verification benchmark (MVR-50, 50 tasks) under paired static and rushing Byzantine attacks, H-CSC commits 0.90/0.92 with honest-reference-invalid rates of 0.02/0.00, statistically matching a strong certificate-emitting verdict-only baseline. Unlike that baseline, H-CSC also emits an embedding-backed semantic_commit digest on 74%/72% of rounds, supplying typed provenance. A strict-semantic ablation commits only 0.54/0.48, showing the verdict-level fallback is necessary for coverage (+0.36/+0.44) at the same <=0.04 safety floor; a 100-task cross-model check across four LLMs preserves invalid_hmaj within 0.00 to 0.03.

Problem

• Byzantine-resilient
• LLM-agent collaboration
• semantic commitment
• finality control
• typed abort

Innovation

• Hierarchical Certified Semantic Commitment
• Byzantine-resilient LLM collaboration
• typed finality
• semantic_commit
• embedding-based consensus