analiza aiqbench
The paper tackles a practical tension in retrieval-augmented LLMs: sending knowledge-graph evidence to a remote model can leak sensitive structure, and simply sending less hurts multi-hop reasoning. Ctrl-G's local sanitization, controller, and experience memory are a concrete design for limiting both per-call and cumulative disclosure. Readers should treat the reported accuracy gains (up to 13.8 points on five of six benchmarks) as benchmark-specific results, and check how privacy is measured before assuming the protection holds in deployment.
Ctrl-G: Controlling Structural and Interaction Exposure in Knowledge Graph Augmented LLM Reasoning
arXiv:2601.08739v2 Announce Type: replace Abstract: Knowledge graphs (KGs) provide structured evidence for Web question answering with large language models (LLMs), but sending retrieved evidence to a remote LLM can expose sensitive graph information. Existing privacy treatments use entity anonymization and relation-centric abstraction to preserve reasoning cues, but transmitted topology remains visible and disclosure accumulates across repeated calls. Distinctive topology can support re-identification of masked entities (structural exposure), while successive calls accumulate additional paths and exploration traces into a broader view of the KG (interaction exposure). Simply sending less evidence is insufficient because restricted, stepwise views may omit the bridge paths required for multi-hop reasoning. To address this, we propose Ctrl-G, to the best of our knowledge the first KG-RAG framework to jointly control structural and interaction exposure to a remote LLM while supporting accurate multi-hop reasoning. To control structural exposure, Ctrl-G locally constructs query-conditioned sanitized views through structural aggregation and pruning, limiting the graph evidence transmitted to a remote LLM. A hierarchical controller couples remote planning with local connected-path exploration across topic entities, using question indicators and retrieved experience to prune candidates while grounding and verifying retained evidence against the original KG. To control interaction exposure, local experience memory reuses abstract, verified reasoning traces to guide exploration modes and depths and gate remote LLM calls, reducing repeated disclosure across questions. Ctrl-G achieves the highest accuracy among the compared methods on five of six benchmarks, exceeding the strongest baseline by up to 13.8 points across model configurations.
Abstract:Knowledge graphs (KGs) provide structured evidence for Web question answering with large language models (LLMs), but sending retrieved evidence to a remote LLM can expose sensitive graph information. Existing privacy treatments use entity anonymization and relation-centric abstraction to preserve reasoning cues, but transmitted topology remains visible and disclosure accumulates across repeated calls. Distinctive topology can support re-identification of masked entities (structural exposure), while successive calls accumulate additional paths and exploration traces into a broader view of the KG (interaction exposure). Simply sending less evidence is insufficient because restricted, stepwise views may omit the bridge paths required for multi-hop reasoning. To address this, we propose Ctrl-G, to the best of our knowledge the first KG-RAG framework to jointly control structural and interaction exposure to a remote LLM while supporting accurate multi-hop reasoning. To control structural exposure, Ctrl-G locally constructs query-conditioned sanitized views through structural aggregation and pruning, limiting the graph evidence transmitted to a remote LLM. A hierarchical controller couples remote planning with local connected-path exploration across topic entities, using question indicators and retrieved experience to prune candidates while grounding and verifying retained evidence against the original KG. To control interaction exposure, local experience memory reuses abstract, verified reasoning traces to guide exploration modes and depths and gate remote LLM calls, reducing repeated disclosure across questions. Ctrl-G achieves the highest accuracy among the compared methods on five of six benchmarks, exceeding the strongest baseline by up to 13.8 points across model configurations.
Computation and Language (cs.CL)
arXiv:2601.08739 [cs.CL]
(or arXiv:2601.08739v2 [cs.CL] for this version)
https://doi.org/10.48550/arXiv.2601.08739 Focus to learn more arXiv-issued DOI via DataCite
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Text integral sub licența deschisă a publicației · 334 cuvinte
Subiecte conexe
Micro-World 2: Streaming Real-time Interactive World Model
AMD · 21 oct. 2026
Composite-Gradient Learning for Shared Control Authority Between Deep Reinforcement Learning and Model Predictive Control
arXiv · 17 sept. 2026
A Unified Control Architecture for Macro-Micro Manipulation using a Active Remote Center of Compliance for Manufacturing Applications
arXiv · 16 iun. 2026
Interaction Dynamics Modeling and Predictive Control for Safe Steerable Catheter--Tissue Interaction
arXiv · 24 iul. 2026