{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:BIYQTX5YZHZJWTUO2RF27FGZNI","short_pith_number":"pith:BIYQTX5Y","schema_version":"1.0","canonical_sha256":"0a3109dfb8c9f29b4e8ed44baf94d96a03915661c6063b5e143ce531c98a1648","source":{"kind":"arxiv","id":"2606.04051","version":1},"attestation_state":"computed","paper":{"title":"RUBAS: Rubric-Based Reinforcement Learning for Agent Safety","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","cs.CR"],"primary_cat":"cs.LG","authors_text":"Fei Mi, Hongning Wang, Minlie Huang, Qinglin Su, Qi Zhu, Shiyao Cui, Xian Qi Loye, Zhexin Zhang","submitted_at":"2026-06-02T09:02:14Z","abstract_excerpt":"The evolution of LLMs into tool-enabled agents creates a new class of safety challenges associated with real-world execution rather than simple text generation. Existing alignment methods often rely on coarse refusal signals or static supervision, making it difficult to balance safety with useful tool execution across diverse agentic risks. We introduce RUBAS, a rubric-based reinforcement learning framework for agent safety. RUBAS decomposes agent behavior into four dimensions: tool-use safety, argument safety, response safety, and helpfulness. These structured rubrics provide fine-grained and"},"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":"2606.04051","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2026-06-02T09:02:14Z","cross_cats_sorted":["cs.AI","cs.CR"],"title_canon_sha256":"a7aa620a80038e051a24c49d7937f3ccbadd9a2bb41f0e732151073a5838f0b2","abstract_canon_sha256":"ce83c7d38e399a8b6934ed2ef28ddbb4cd03f62894daf3d6e1d0029ba42f8191"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-06-04T00:06:45.304962Z","signature_b64":"ZMJwWBaq0orPrzTdXmVhg8MPoFKm++Yhuvh3yIEd8+ENiNYKOKFLPGgORZ7pv7m+k+hNkB70Yw0EzODY+FKeDw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"0a3109dfb8c9f29b4e8ed44baf94d96a03915661c6063b5e143ce531c98a1648","last_reissued_at":"2026-06-04T00:06:45.304547Z","signature_status":"signed_v1","first_computed_at":"2026-06-04T00:06:45.304547Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"RUBAS: Rubric-Based Reinforcement Learning for Agent Safety","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","cs.CR"],"primary_cat":"cs.LG","authors_text":"Fei Mi, Hongning Wang, Minlie Huang, Qinglin Su, Qi Zhu, Shiyao Cui, Xian Qi Loye, Zhexin Zhang","submitted_at":"2026-06-02T09:02:14Z","abstract_excerpt":"The evolution of LLMs into tool-enabled agents creates a new class of safety challenges associated with real-world execution rather than simple text generation. Existing alignment methods often rely on coarse refusal signals or static supervision, making it difficult to balance safety with useful tool execution across diverse agentic risks. We introduce RUBAS, a rubric-based reinforcement learning framework for agent safety. RUBAS decomposes agent behavior into four dimensions: tool-use safety, argument safety, response safety, and helpfulness. These structured rubrics provide fine-grained and"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2606.04051","kind":"arxiv","version":1},"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/2606.04051/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":"2606.04051","created_at":"2026-06-04T00:06:45.304605+00:00"},{"alias_kind":"arxiv_version","alias_value":"2606.04051v1","created_at":"2026-06-04T00:06:45.304605+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2606.04051","created_at":"2026-06-04T00:06:45.304605+00:00"},{"alias_kind":"pith_short_12","alias_value":"BIYQTX5YZHZJ","created_at":"2026-06-04T00:06:45.304605+00:00"},{"alias_kind":"pith_short_16","alias_value":"BIYQTX5YZHZJWTUO","created_at":"2026-06-04T00:06:45.304605+00:00"},{"alias_kind":"pith_short_8","alias_value":"BIYQTX5Y","created_at":"2026-06-04T00:06:45.304605+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":0,"internal_anchor_count":0,"sample":[]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/BIYQTX5YZHZJWTUO2RF27FGZNI","json":"https://pith.science/pith/BIYQTX5YZHZJWTUO2RF27FGZNI.json","graph_json":"https://pith.science/api/pith-number/BIYQTX5YZHZJWTUO2RF27FGZNI/graph.json","events_json":"https://pith.science/api/pith-number/BIYQTX5YZHZJWTUO2RF27FGZNI/events.json","paper":"https://pith.science/paper/BIYQTX5Y"},"agent_actions":{"view_html":"https://pith.science/pith/BIYQTX5YZHZJWTUO2RF27FGZNI","download_json":"https://pith.science/pith/BIYQTX5YZHZJWTUO2RF27FGZNI.json","view_paper":"https://pith.science/paper/BIYQTX5Y","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2606.04051&json=true","fetch_graph":"https://pith.science/api/pith-number/BIYQTX5YZHZJWTUO2RF27FGZNI/graph.json","fetch_events":"https://pith.science/api/pith-number/BIYQTX5YZHZJWTUO2RF27FGZNI/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/BIYQTX5YZHZJWTUO2RF27FGZNI/action/timestamp_anchor","attest_storage":"https://pith.science/pith/BIYQTX5YZHZJWTUO2RF27FGZNI/action/storage_attestation","attest_author":"https://pith.science/pith/BIYQTX5YZHZJWTUO2RF27FGZNI/action/author_attestation","sign_citation":"https://pith.science/pith/BIYQTX5YZHZJWTUO2RF27FGZNI/action/citation_signature","submit_replication":"https://pith.science/pith/BIYQTX5YZHZJWTUO2RF27FGZNI/action/replication_record"}},"created_at":"2026-06-04T00:06:45.304605+00:00","updated_at":"2026-06-04T00:06:45.304605+00:00"}