ARBITRA
A document-level ATT&CK extraction framework for cyber threat intelligence. Multiple expert LLM roles propose candidate entities, and each candidate is verified through a Quantitative Bipolar Argumentation Framework (QBAF), combining document-level context utilization with high-precision extraction.
Timeline (last 90 days)
ARBITRA is a new framework that achieves robust document-level ATT&CK extraction through multi-expert argumentation.
The ARBITRA workflow is divided into two stages: proposal and verification.
In the proposal stage, ARBITRA has multiple role-specialized LLM experts each identify possible ATT&CK entities from the full report, forming a candidate set.
In the verification stage, ARBITRA adopts a quantitative bipolar argumentation framework (QBAF) to quantitatively weigh supporting and opposing evidence for each candidate entity based on document evidence and reasoning grounds.
On the ATT&CK extraction benchmark, ARBITRA achieves a stronger precision-oriented operating point than sentence-level pipelines, recent LLM-based extraction systems, and frontier models including GPT-5.5.
Under adversarial perturbations simulating conflicting TTP statements, cross-report contamination, and semantic poisoning, ARBITRA limits average precision degradation to about half that of the frontier-model baselines.
ARBITRA's multi-expert argumentation approach provides a reference for other structured information extraction tasks; when a single model struggles to balance contextual completeness and extraction precision, introducing multi-role collaboration and formal argumentation verification may be an effective path.
All Facts (7)
ARBITRA is a new framework that achieves robust document-level ATT&CK extraction through multi-expert argumentation.
50%UnverifiedThe ARBITRA workflow is divided into two stages: proposal and verification.
50%UnverifiedIn the proposal stage, ARBITRA has multiple role-specialized LLM experts each identify possible ATT&CK entities from the full report, forming a candidate set.
50%UnverifiedIn the verification stage, ARBITRA adopts a quantitative bipolar argumentation framework (QBAF) to quantitatively weigh supporting and opposing evidence for each candidate entity based on document evidence and reasoning grounds.
50%UnverifiedOn the ATT&CK extraction benchmark, ARBITRA achieves a stronger precision-oriented operating point than sentence-level pipelines, recent LLM-based extraction systems, and frontier models including GPT-5.5.
50%UnverifiedUnder adversarial perturbations simulating conflicting TTP statements, cross-report contamination, and semantic poisoning, ARBITRA limits average precision degradation to about half that of the frontier-model baselines.
50%UnverifiedARBITRA's multi-expert argumentation approach provides a reference for other structured information extraction tasks; when a single model struggles to balance contextual completeness and extraction precision, introducing multi-role collaboration and formal argumentation verification may be an effective path.
50%