Evidence-Bound Autonomous Research (EviBound): A Governance Framework for Eliminating False Claims

2025/10/28 by Ruiying Chen, Chen, Ruiying · 1 citation
Biochemistry, Genetics and Molecular Biology · Computer Science · Decision Sciences · #Artificial Intelligence (cs.AI) #Biomedical Text Mining and Ontologies #FOS: Computer and information sciences #Research Data Management Practices #Scientific Computing and Data Management

paper · pdf · doi:10.48550/arxiv.2511.05524

openalex publication_date 2025/10/28 · openalex created_date 2025/11/12 · openalex updated_date 2026/07/28

Abstract

LLM-based autonomous research agents report false claims: tasks marked "complete" despite missing artifacts, contradictory metrics, or failed executions. EviBound is an evidence-bound execution framework that eliminates false claims through dual governance gates requiring machine-checkable evidence. Two complementary gates enforce evidence requirements. The pre-execution Approval Gate validates acceptance criteria schemas before code runs, catching structural violations proactively. The post-execution Verification Gate validates artifacts via MLflow API queries (with recursive path checking) and optionally validates metrics when specified by acceptance criteria. Claims propagate only when backed by a queryable run ID, required artifacts, and FINISHED status. Bounded, confidence-gated retries (typically 1-2 attempts) recover from transient failures without unbounded loops. The framework was evaluated on 8 benchmark tasks spanning infrastructure validation, ML capabilities, and governance stress tests. Baseline A (Prompt-Level Only) yields 100% hallucination (8/8 claimed, 0/8 verified). Baseline B (Verification-Only) reduces hallucination to 25% (2/8 fail verification). EviBound (Dual Gates) achieves 0% hallucination: 7/8 tasks verified and 1 task correctly blocked at the approval gate, all with only approximately 8.3% execution overhead. This package includes execution trajectories, MLflow run IDs for all verified tasks, and a 4-step verification protocol. Research integrity is an architectural property, achieved through governance gates rather than emergent from model scale.

Citations

• MLR-Copilot: Autonomous Machine Learning Research based on Large Language Models Agents
• The AI Scientist: Towards Fully Automated Open-Ended Scientific Discovery
• A Comprehensive Survey of Hallucination Mitigation Techniques in Large Language Models
• Towards Mitigating Hallucination in Large Language Models via Self-Reflection
• Counterfactually Auditable Lifecycle Certification for Autonomous Agents
• FActScore: Fine-grained Atomic Evaluation of Factual Precision in Long Form Text Generation
• Reflexion: Language Agents with Verbal Reinforcement Learning
• SelfCheckGPT: Zero-Resource Black-Box Hallucination Detection for Generative Large Language Models
• Toolformer: Language Models Can Teach Themselves to Use Tools
• PAL: Program-aided Language Models
• Distributing Accountability, Not Capability: Phase Separation and the LLM Workflow Quadrant in Autonomous AI Agent Architectures
• MRKL Systems: A modular, neuro-symbolic architecture that combines large language models, external knowledge sources and discrete reasoning

Cited by

• Safety, or Just Capability? A Validity Audit of Agent-Safety Benchmarks

Related

• Proof-or-Stop: Don't Trust the Agent, Trust the Evidence -- Loop Engineering for Verifiable Evidence-Gated Lifecycle Control
• EvalSafetyGap: A Hybrid Survey and Conceptual Framework for LLM Evaluation-Safety Failures
• HALLMARK: Diagnosing Three Failure Modes in LLM Citation Verifiers
• DRNOISE: Benchmarking Deep Research Agents in Misleading Evidence Environments
• Evaluating and Guarding Citation Faithfulness in Agentic Scientific Synthesis