Correct Answers, Unsupported Findings: Evidence Binding in Forensic Reconstruction of LLM Agent Logs

cs.CR, cs.AI

Submitted: 2026-10-07

Updated: 2026-10-07

Comments: 11 pages, 2 figures

License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/

Sources

• Reasoning Models Don't Always Say What They Think
• Evaluating and Guarding Citation Faithfulness in Agentic Scientific Synthesis
• Auditing Provenance Sensitivity in LLM Agent Action Selection
• AutoDojo: A Generative Benchmark for Evaluating Prompt Injection Defenses in LLM Agents
• Observability for Delegated Execution in Agentic AI Systems
• Aligning Provenance with Authorization: A Dual-Graph Defense for LLM Agents

Related papers

• SoK: AI-Augmented Binary Reversing
• Relaxed Sender Anonymity for CBDC Interbank Settlement: A Zero-Knowledge Approach on Permissioned EVM
• Calibration-Family Overfit: Why Trusted Sabotage Monitors Don't Transfer Across Lineages
• Efficient Fuzzy PSI under One-Sided Assumptions
• Sealing the Audit-Runtime Gap for LLM Skills
• Token Composition: A Graph Based on EVM Logs