{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:WGOLOEUSJNVKEN5XFPYXL7ZXIW","short_pith_number":"pith:WGOLOEUS","schema_version":"1.0","canonical_sha256":"b19cb712924b6aa237b72bf175ff3745a8b82b86c339a3930063004e9862c366","source":{"kind":"arxiv","id":"2605.03409","version":2},"attestation_state":"computed","paper":{"title":"Robust Agent Compensation (RAC): Teaching AI Agents to Compensate","license":"http://creativecommons.org/licenses/by/4.0/","headline":"AI agents can recover from failures using a log-based safety net added through existing framework extensions without rewriting their code.","cross_cats":[],"primary_cat":"cs.AI","authors_text":"Frank Leymann, Kaviru Hapuarachchi, Rania Khalaf, Srinath Perera","submitted_at":"2026-05-05T06:27:34Z","abstract_excerpt":"We present Robust Agent Compensation (RAC), a log-based recovery paradigm (providing a safety net) implemented through an architectural extension that can be applied to most Agent frameworks to support reliable executions (avoiding unintended side effects). Users can choose to enable RAC without changing their current agent code (e.g., LangGraph agents). The proposed approach can be implemented in most existing agent frameworks via their existing extension points. We present an implementation based on LangChain, demonstrate its viability through the $\\tau$-bench and REALM-Bench, and show that "},"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":"2605.03409","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.AI","submitted_at":"2026-05-05T06:27:34Z","cross_cats_sorted":[],"title_canon_sha256":"1d1018d895408d85bad66e3ba6e41fb678c45b5138a0a5d3bbba783599768827","abstract_canon_sha256":"13d61640aab8b9c27af74e7ef88e875cb2ab374847b01ecc1e84c379476165d7"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-20T00:04:33.876245Z","signature_b64":"ydXtpZXhU9vWUB8OAGSOCO0ZoTSmAFdZNrwzu1j9x1KDhMwdoPO86mantpERf/uWkPXKcxWVMQXvlejfz8DkBA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"b19cb712924b6aa237b72bf175ff3745a8b82b86c339a3930063004e9862c366","last_reissued_at":"2026-05-20T00:04:33.875305Z","signature_status":"signed_v1","first_computed_at":"2026-05-20T00:04:33.875305Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Robust Agent Compensation (RAC): Teaching AI Agents to Compensate","license":"http://creativecommons.org/licenses/by/4.0/","headline":"AI agents can recover from failures using a log-based safety net added through existing framework extensions without rewriting their code.","cross_cats":[],"primary_cat":"cs.AI","authors_text":"Frank Leymann, Kaviru Hapuarachchi, Rania Khalaf, Srinath Perera","submitted_at":"2026-05-05T06:27:34Z","abstract_excerpt":"We present Robust Agent Compensation (RAC), a log-based recovery paradigm (providing a safety net) implemented through an architectural extension that can be applied to most Agent frameworks to support reliable executions (avoiding unintended side effects). Users can choose to enable RAC without changing their current agent code (e.g., LangGraph agents). The proposed approach can be implemented in most existing agent frameworks via their existing extension points. We present an implementation based on LangChain, demonstrate its viability through the $\\tau$-bench and REALM-Bench, and show that "},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"We present an implementation based on LangChain, demonstrate its viability through the τ-bench and REALM-Bench, and show that when solving complex problems, RAC is 1.5-8X or more better in both latency and token economy compared to state-of-the-art LLM-based recovery approaches.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"That the log-based recovery can be implemented via existing extension points without changing current agent code and that it effectively avoids unintended side effects.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"RAC adds a log-based safety net to AI agents via framework extensions, delivering 1.5-8X better latency and token use than LLM-based recovery on complex problems in τ-bench and REALM-Bench.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"AI agents can recover from failures using a log-based safety net added through existing framework extensions without rewriting their code.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"a4c80f9efe86d5a00049ba3a95cb21d8da9ae53370cbc9dee582ae640b74f74b"},"source":{"id":"2605.03409","kind":"arxiv","version":2},"verdict":{"id":"d75e52d0-81e7-4787-a939-3d2b4d213c1c","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-07T16:49:15.217035Z","strongest_claim":"We present an implementation based on LangChain, demonstrate its viability through the τ-bench and REALM-Bench, and show that when solving complex problems, RAC is 1.5-8X or more better in both latency and token economy compared to state-of-the-art LLM-based recovery approaches.","one_line_summary":"RAC adds a log-based safety net to AI agents via framework extensions, delivering 1.5-8X better latency and token use than LLM-based recovery on complex problems in τ-bench and REALM-Bench.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"That the log-based recovery can be implemented via existing extension points without changing current agent code and that it effectively avoids unintended side effects.","pith_extraction_headline":"AI agents can recover from failures using a log-based safety net added through existing framework extensions without rewriting their code."},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2605.03409/integrity.json","findings":[],"available":true,"detectors_run":[{"name":"doi_compliance","ran_at":"2026-05-19T15:26:22.369536Z","status":"completed","version":"1.0.0","findings_count":0}],"snapshot_sha256":"cc0b42458a06ee94f017f0e4a1cdb826a76eb90846be02c9ec4b491c1cd46bfe"},"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":"2605.03409","created_at":"2026-05-20T00:04:33.875437+00:00"},{"alias_kind":"arxiv_version","alias_value":"2605.03409v2","created_at":"2026-05-20T00:04:33.875437+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.03409","created_at":"2026-05-20T00:04:33.875437+00:00"},{"alias_kind":"pith_short_12","alias_value":"WGOLOEUSJNVK","created_at":"2026-05-20T00:04:33.875437+00:00"},{"alias_kind":"pith_short_16","alias_value":"WGOLOEUSJNVKEN5X","created_at":"2026-05-20T00:04:33.875437+00:00"},{"alias_kind":"pith_short_8","alias_value":"WGOLOEUS","created_at":"2026-05-20T00:04:33.875437+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/WGOLOEUSJNVKEN5XFPYXL7ZXIW","json":"https://pith.science/pith/WGOLOEUSJNVKEN5XFPYXL7ZXIW.json","graph_json":"https://pith.science/api/pith-number/WGOLOEUSJNVKEN5XFPYXL7ZXIW/graph.json","events_json":"https://pith.science/api/pith-number/WGOLOEUSJNVKEN5XFPYXL7ZXIW/events.json","paper":"https://pith.science/paper/WGOLOEUS"},"agent_actions":{"view_html":"https://pith.science/pith/WGOLOEUSJNVKEN5XFPYXL7ZXIW","download_json":"https://pith.science/pith/WGOLOEUSJNVKEN5XFPYXL7ZXIW.json","view_paper":"https://pith.science/paper/WGOLOEUS","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2605.03409&json=true","fetch_graph":"https://pith.science/api/pith-number/WGOLOEUSJNVKEN5XFPYXL7ZXIW/graph.json","fetch_events":"https://pith.science/api/pith-number/WGOLOEUSJNVKEN5XFPYXL7ZXIW/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/WGOLOEUSJNVKEN5XFPYXL7ZXIW/action/timestamp_anchor","attest_storage":"https://pith.science/pith/WGOLOEUSJNVKEN5XFPYXL7ZXIW/action/storage_attestation","attest_author":"https://pith.science/pith/WGOLOEUSJNVKEN5XFPYXL7ZXIW/action/author_attestation","sign_citation":"https://pith.science/pith/WGOLOEUSJNVKEN5XFPYXL7ZXIW/action/citation_signature","submit_replication":"https://pith.science/pith/WGOLOEUSJNVKEN5XFPYXL7ZXIW/action/replication_record"}},"created_at":"2026-05-20T00:04:33.875437+00:00","updated_at":"2026-05-20T00:04:33.875437+00:00"}