{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2019:HAV22UYRN4KIILGKLSOQGRQQOY","short_pith_number":"pith:HAV22UYR","schema_version":"1.0","canonical_sha256":"382bad53116f14842cca5c9d03461076054429c093ab72b1e2d91046b39fb8e1","source":{"kind":"arxiv","id":"1905.10615","version":3},"attestation_state":"computed","paper":{"title":"Adversarial Policies: Attacking Deep Reinforcement Learning","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","cs.CR","stat.ML"],"primary_cat":"cs.LG","authors_text":"Adam Gleave, Cody Wild, Michael Dennis, Neel Kant, Sergey Levine, Stuart Russell","submitted_at":"2019-05-25T15:23:19Z","abstract_excerpt":"Deep reinforcement learning (RL) policies are known to be vulnerable to adversarial perturbations to their observations, similar to adversarial examples for classifiers. However, an attacker is not usually able to directly modify another agent's observations. This might lead one to wonder: is it possible to attack an RL agent simply by choosing an adversarial policy acting in a multi-agent environment so as to create natural observations that are adversarial? We demonstrate the existence of adversarial policies in zero-sum games between simulated humanoid robots with proprioceptive observation"},"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":"1905.10615","kind":"arxiv","version":3},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2019-05-25T15:23:19Z","cross_cats_sorted":["cs.AI","cs.CR","stat.ML"],"title_canon_sha256":"17fd1a4bb0ee0b5f9d1445bb830d0e57455885be4bf7144b6a8f22807244f708","abstract_canon_sha256":"a93b4d6ace5a99c500440cb0e7a411cd65e7d46369020629d3504e4caedb5f18"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T02:07:25.878525Z","signature_b64":"Uyrf+IW2y35L4+3HCo1VEBY3QP3NURDPBPB6HnopI00MG/fq4IajhRMeai46kACBDjXyRIj3EqamK2D/0faADw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"382bad53116f14842cca5c9d03461076054429c093ab72b1e2d91046b39fb8e1","last_reissued_at":"2026-07-05T02:07:25.878041Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T02:07:25.878041Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Adversarial Policies: Attacking Deep Reinforcement Learning","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","cs.CR","stat.ML"],"primary_cat":"cs.LG","authors_text":"Adam Gleave, Cody Wild, Michael Dennis, Neel Kant, Sergey Levine, Stuart Russell","submitted_at":"2019-05-25T15:23:19Z","abstract_excerpt":"Deep reinforcement learning (RL) policies are known to be vulnerable to adversarial perturbations to their observations, similar to adversarial examples for classifiers. However, an attacker is not usually able to directly modify another agent's observations. This might lead one to wonder: is it possible to attack an RL agent simply by choosing an adversarial policy acting in a multi-agent environment so as to create natural observations that are adversarial? We demonstrate the existence of adversarial policies in zero-sum games between simulated humanoid robots with proprioceptive observation"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"1905.10615","kind":"arxiv","version":3},"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/1905.10615/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":"1905.10615","created_at":"2026-07-05T02:07:25.878102+00:00"},{"alias_kind":"arxiv_version","alias_value":"1905.10615v3","created_at":"2026-07-05T02:07:25.878102+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.1905.10615","created_at":"2026-07-05T02:07:25.878102+00:00"},{"alias_kind":"pith_short_12","alias_value":"HAV22UYRN4KI","created_at":"2026-07-05T02:07:25.878102+00:00"},{"alias_kind":"pith_short_16","alias_value":"HAV22UYRN4KIILGK","created_at":"2026-07-05T02:07:25.878102+00:00"},{"alias_kind":"pith_short_8","alias_value":"HAV22UYR","created_at":"2026-07-05T02:07:25.878102+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":7,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.30807","citing_title":"Off the Rails: Hijacking the Scoring Head in Generative End-to-End Driving Planners with Safety-Violating Adversarial Perturbations","ref_index":14,"is_internal_anchor":false},{"citing_arxiv_id":"2605.23027","citing_title":"PIMbot: A Self-Adaptive Attack Framework for Adversarial Manipulation of Multi-Robot Reinforcement Learning","ref_index":29,"is_internal_anchor":false},{"citing_arxiv_id":"2502.03698","citing_title":"How Vulnerable Is My Learned Policy? Universal Adversarial Perturbation Attacks On Modern Behavior Cloning Policies","ref_index":24,"is_internal_anchor":false},{"citing_arxiv_id":"2509.15103","citing_title":"Vulnerable Agent Identification in Large-Scale Multi-Agent Reinforcement Learning","ref_index":14,"is_internal_anchor":false},{"citing_arxiv_id":"2604.10974","citing_title":"Robust Adversarial Policy Optimization Under Dynamics Uncertainty","ref_index":4,"is_internal_anchor":false},{"citing_arxiv_id":"2604.04733","citing_title":"Discovering Failure Modes in Vision-Language Models using RL","ref_index":8,"is_internal_anchor":false},{"citing_arxiv_id":"2605.02495","citing_title":"Efficient Preference Poisoning Attack on Offline RLHF","ref_index":130,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/HAV22UYRN4KIILGKLSOQGRQQOY","json":"https://pith.science/pith/HAV22UYRN4KIILGKLSOQGRQQOY.json","graph_json":"https://pith.science/api/pith-number/HAV22UYRN4KIILGKLSOQGRQQOY/graph.json","events_json":"https://pith.science/api/pith-number/HAV22UYRN4KIILGKLSOQGRQQOY/events.json","paper":"https://pith.science/paper/HAV22UYR"},"agent_actions":{"view_html":"https://pith.science/pith/HAV22UYRN4KIILGKLSOQGRQQOY","download_json":"https://pith.science/pith/HAV22UYRN4KIILGKLSOQGRQQOY.json","view_paper":"https://pith.science/paper/HAV22UYR","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=1905.10615&json=true","fetch_graph":"https://pith.science/api/pith-number/HAV22UYRN4KIILGKLSOQGRQQOY/graph.json","fetch_events":"https://pith.science/api/pith-number/HAV22UYRN4KIILGKLSOQGRQQOY/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/HAV22UYRN4KIILGKLSOQGRQQOY/action/timestamp_anchor","attest_storage":"https://pith.science/pith/HAV22UYRN4KIILGKLSOQGRQQOY/action/storage_attestation","attest_author":"https://pith.science/pith/HAV22UYRN4KIILGKLSOQGRQQOY/action/author_attestation","sign_citation":"https://pith.science/pith/HAV22UYRN4KIILGKLSOQGRQQOY/action/citation_signature","submit_replication":"https://pith.science/pith/HAV22UYRN4KIILGKLSOQGRQQOY/action/replication_record"}},"created_at":"2026-07-05T02:07:25.878102+00:00","updated_at":"2026-07-05T02:07:25.878102+00:00"}