{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:I7G5YX247BSO7XCSY5IVK6WQMN","short_pith_number":"pith:I7G5YX24","schema_version":"1.0","canonical_sha256":"47cddc5f5cf864efdc52c751557ad0634070faebf830f1304facc23c62840c02","source":{"kind":"arxiv","id":"2508.00500","version":4},"attestation_state":"computed","paper":{"title":"ProbGuard: Proactive Runtime Monitoring for LLM Agent Safety via Probabilistic Prediction","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.SE"],"primary_cat":"cs.AI","authors_text":"Christopher M. Poskitt, Haoyu Wang, Jiali Wei, Jun Sun","submitted_at":"2025-08-01T10:24:47Z","abstract_excerpt":"Large Language Model (LLM) agents increasingly operate across domains such as robotics, virtual assistants, and web automation. However, their stochastic decision-making introduces safety risks that are difficult to anticipate during execution. Existing runtime monitoring frameworks, such as AgentSpec, primarily rely on reactive safety rules that detect violations only when unsafe behavior is imminent or has already occurred, limiting their ability to handle long-horizon dependencies. We present ProbGuard, a proactive runtime monitoring framework for LLM agents that anticipates safety violatio"},"verification_status":{"content_addressed":true,"pith_receipt":true,"author_attested":false,"weak_author_claims":0,"strong_author_claims":0,"externally_anchored":false,"storage_verified":false,"citation_signatures":0,"replication_records":0,"graph_snapshot":true,"references_resolved":false,"formal_links_present":false},"canonical_record":{"source":{"id":"2508.00500","kind":"arxiv","version":4},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.AI","submitted_at":"2025-08-01T10:24:47Z","cross_cats_sorted":["cs.SE"],"title_canon_sha256":"454654a232c045bf25e7d80df69e3fdb0c554d569c5abedab51e3d0215e3b462","abstract_canon_sha256":"8ce97c36099b9490aa5a81dc782662bb774007cf6e32bd72a25aa4273e98f604"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-08-04T01:57:50.809441Z","signature_b64":"Ih9lOoPsFiBkNtBNiA9DPlH32ADhlPDt/Qhqh6h+lssaHhKyc3DLOrBZ7KQ4niRmssZRO14hF5D4SUKj6AyIBw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"47cddc5f5cf864efdc52c751557ad0634070faebf830f1304facc23c62840c02","last_reissued_at":"2026-08-04T01:57:50.807867Z","signature_status":"signed_v1","first_computed_at":"2026-08-04T01:57:50.807867Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"ProbGuard: Proactive Runtime Monitoring for LLM Agent Safety via Probabilistic Prediction","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.SE"],"primary_cat":"cs.AI","authors_text":"Christopher M. Poskitt, Haoyu Wang, Jiali Wei, Jun Sun","submitted_at":"2025-08-01T10:24:47Z","abstract_excerpt":"Large Language Model (LLM) agents increasingly operate across domains such as robotics, virtual assistants, and web automation. However, their stochastic decision-making introduces safety risks that are difficult to anticipate during execution. Existing runtime monitoring frameworks, such as AgentSpec, primarily rely on reactive safety rules that detect violations only when unsafe behavior is imminent or has already occurred, limiting their ability to handle long-horizon dependencies. We present ProbGuard, a proactive runtime monitoring framework for LLM agents that anticipates safety violatio"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2508.00500","kind":"arxiv","version":4},"verdict":{"id":null,"model_set":{},"created_at":null,"strongest_claim":"","one_line_summary":"","pipeline_version":null,"weakest_assumption":"","pith_extraction_headline":""},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2508.00500/integrity.json","findings":[],"available":true,"detectors_run":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938"},"references":{"count":0,"sample":[],"resolved_work":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57","internal_anchors":0},"formal_canon":{"evidence_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"},"aliases":[{"alias_kind":"arxiv","alias_value":"2508.00500","created_at":"2026-08-04T01:57:50.809164+00:00"},{"alias_kind":"arxiv_version","alias_value":"2508.00500v4","created_at":"2026-08-04T01:57:50.809164+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2508.00500","created_at":"2026-08-04T01:57:50.809164+00:00"},{"alias_kind":"pith_short_12","alias_value":"I7G5YX247BSO","created_at":"2026-08-04T01:57:50.809164+00:00"},{"alias_kind":"pith_short_16","alias_value":"I7G5YX247BSO7XCS","created_at":"2026-08-04T01:57:50.809164+00:00"},{"alias_kind":"pith_short_8","alias_value":"I7G5YX24","created_at":"2026-08-04T01:57:50.809164+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":24,"internal_anchor_count":24,"sample":[{"citing_arxiv_id":"2606.24245","citing_title":"AutoSpec: Safety Rule Evolution for LLM Agents via Inductive Logic Programming","ref_index":34,"is_internal_anchor":true},{"citing_arxiv_id":"2606.19386","citing_title":"Bistable by Construction: Wall-Clock-Calibrated State Monitors Have No Moment-Detection Regime at Agent Cadence","ref_index":20,"is_internal_anchor":true},{"citing_arxiv_id":"2607.02510","citing_title":"Online Safety Monitoring for LLMs","ref_index":28,"is_internal_anchor":true},{"citing_arxiv_id":"2606.04296","citing_title":"The Saturation Trap and the Subjectivity of Intervention Timing: Why Affect-Based Triggers and LLM Judges Fail to Time Interventions on Autonomous Agents","ref_index":20,"is_internal_anchor":true},{"citing_arxiv_id":"2606.27570","citing_title":"Auditing AI Investment Recommendations as Executable Actions","ref_index":32,"is_internal_anchor":true},{"citing_arxiv_id":"2606.28061","citing_title":"ToolPrivacyBench: Benchmarking Purpose-Bound Privacy in Tool-Using LLM Agents","ref_index":28,"is_internal_anchor":true},{"citing_arxiv_id":"2604.07833","citing_title":"Harnessing Embodied Agents: Runtime Governance for Policy-Constrained Execution","ref_index":52,"is_internal_anchor":true},{"citing_arxiv_id":"2605.16471","citing_title":"From AI-Generated Content to Agentic Action: Security and Safety Threats in Generative AI","ref_index":135,"is_internal_anchor":true},{"citing_arxiv_id":"2605.18672","citing_title":"Position: A Three-Layer Probabilistic Assume-Guarantee Architecture Is Structurally Required for Safe LLM Agent Deployment","ref_index":61,"is_internal_anchor":true},{"citing_arxiv_id":"2605.14859","citing_title":"Do Coding Agents Understand Least-Privilege Authorization?","ref_index":40,"is_internal_anchor":true},{"citing_arxiv_id":"2510.23883","citing_title":"Agentic AI Security: Threats, Defenses, Evaluation, and Open Challenges","ref_index":223,"is_internal_anchor":true},{"citing_arxiv_id":"2603.17418","citing_title":"PowerDAG: Supervisory Agentic AI System for Automating Distribution Grid Analysis","ref_index":41,"is_internal_anchor":true},{"citing_arxiv_id":"2604.02022","citing_title":"ATBench: A Diverse and Realistic Agent Trajectory Benchmark for Safety Evaluation and Diagnosis","ref_index":32,"is_internal_anchor":true},{"citing_arxiv_id":"2603.27905","citing_title":"ATLAS-RTC: Closing the Loop on LLM Agent Output with Token-Level Runtime Control","ref_index":11,"is_internal_anchor":true},{"citing_arxiv_id":"2604.02022","citing_title":"ATBench: A Diverse and Realistic Agent Trajectory Benchmark for Safety Evaluation and Diagnosis","ref_index":32,"is_internal_anchor":true},{"citing_arxiv_id":"2605.11874","citing_title":"RecRM-Bench: Benchmarking Multidimensional Reward Modeling for Agentic Recommender Systems","ref_index":37,"is_internal_anchor":true},{"citing_arxiv_id":"2604.24579","citing_title":"Measuring the Unmeasurable: Markov Chain Reliability for LLM Agents","ref_index":21,"is_internal_anchor":true},{"citing_arxiv_id":"2605.06393","citing_title":"Constraining Host-Level Abuse in Self-Hosted Computer-Use Agents via TEE-Backed Isolation","ref_index":41,"is_internal_anchor":true},{"citing_arxiv_id":"2604.22136","citing_title":"Sovereign Agentic Loops: Decoupling AI Reasoning from Execution in Real-World Systems","ref_index":14,"is_internal_anchor":true},{"citing_arxiv_id":"2604.07833","citing_title":"Harnessing Embodied Agents: Runtime Governance for Policy-Constrained Execution","ref_index":52,"is_internal_anchor":true},{"citing_arxiv_id":"2605.06731","citing_title":"When Routine Chats Turn Toxic: Unintended Long-Term State Poisoning in Personalized Agents","ref_index":24,"is_internal_anchor":true},{"citing_arxiv_id":"2605.07935","citing_title":"TraceFix: Repairing Agent Coordination Protocols with TLA+ Counterexamples","ref_index":37,"is_internal_anchor":true},{"citing_arxiv_id":"2604.05119","citing_title":"Governance-Aware Agent Telemetry for Closed-Loop Enforcement in Multi-Agent AI Systems","ref_index":8,"is_internal_anchor":true},{"citing_arxiv_id":"2604.18658","citing_title":"Owner-Harm: A Missing Threat Model for AI Agent Safety","ref_index":7,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/I7G5YX247BSO7XCSY5IVK6WQMN","json":"https://pith.science/pith/I7G5YX247BSO7XCSY5IVK6WQMN.json","graph_json":"https://pith.science/api/pith-number/I7G5YX247BSO7XCSY5IVK6WQMN/graph.json","events_json":"https://pith.science/api/pith-number/I7G5YX247BSO7XCSY5IVK6WQMN/events.json","paper":"https://pith.science/paper/I7G5YX24"},"agent_actions":{"view_html":"https://pith.science/pith/I7G5YX247BSO7XCSY5IVK6WQMN","download_json":"https://pith.science/pith/I7G5YX247BSO7XCSY5IVK6WQMN.json","view_paper":"https://pith.science/paper/I7G5YX24","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2508.00500&json=true","fetch_graph":"https://pith.science/api/pith-number/I7G5YX247BSO7XCSY5IVK6WQMN/graph.json","fetch_events":"https://pith.science/api/pith-number/I7G5YX247BSO7XCSY5IVK6WQMN/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/I7G5YX247BSO7XCSY5IVK6WQMN/action/timestamp_anchor","attest_storage":"https://pith.science/pith/I7G5YX247BSO7XCSY5IVK6WQMN/action/storage_attestation","attest_author":"https://pith.science/pith/I7G5YX247BSO7XCSY5IVK6WQMN/action/author_attestation","sign_citation":"https://pith.science/pith/I7G5YX247BSO7XCSY5IVK6WQMN/action/citation_signature","submit_replication":"https://pith.science/pith/I7G5YX247BSO7XCSY5IVK6WQMN/action/replication_record"}},"created_at":"2026-08-04T01:57:50.809164+00:00","updated_at":"2026-08-04T01:57:50.809164+00:00"}