{"bundle_type":"pith_open_graph_bundle","bundle_version":"1.0","pith_number":"pith:2025:RYIDSCHIDESSAC7JKULDVDQGGJ","short_pith_number":"pith:RYIDSCHI","canonical_record":{"source":{"id":"2505.10670","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.AI","submitted_at":"2025-05-15T19:22:11Z","cross_cats_sorted":["cs.CY","cs.GT"],"title_canon_sha256":"9f361c0122f74023b38c4c684377076ee708d55a27eb943a56315710d9e78a87","abstract_canon_sha256":"4322dcdb486e874f3ed45032c9d4500e08c738e747dcd8c1ab312ee76cbdcbe3"},"schema_version":"1.0"},"canonical_sha256":"8e103908e81925200be955163a8e06325f596760cbc7f03d1568f71c231e0d83","source":{"kind":"arxiv","id":"2505.10670","version":1},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2505.10670","created_at":"2026-07-05T11:04:03Z"},{"alias_kind":"arxiv_version","alias_value":"2505.10670v1","created_at":"2026-07-05T11:04:03Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2505.10670","created_at":"2026-07-05T11:04:03Z"},{"alias_kind":"pith_short_12","alias_value":"RYIDSCHIDESS","created_at":"2026-07-05T11:04:03Z"},{"alias_kind":"pith_short_16","alias_value":"RYIDSCHIDESSAC7J","created_at":"2026-07-05T11:04:03Z"},{"alias_kind":"pith_short_8","alias_value":"RYIDSCHI","created_at":"2026-07-05T11:04:03Z"}],"events":[{"event_type":"record_created","subject_pith_number":"pith:2025:RYIDSCHIDESSAC7JKULDVDQGGJ","target":"record","payload":{"canonical_record":{"source":{"id":"2505.10670","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.AI","submitted_at":"2025-05-15T19:22:11Z","cross_cats_sorted":["cs.CY","cs.GT"],"title_canon_sha256":"9f361c0122f74023b38c4c684377076ee708d55a27eb943a56315710d9e78a87","abstract_canon_sha256":"4322dcdb486e874f3ed45032c9d4500e08c738e747dcd8c1ab312ee76cbdcbe3"},"schema_version":"1.0"},"canonical_sha256":"8e103908e81925200be955163a8e06325f596760cbc7f03d1568f71c231e0d83","receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:04:03.328805Z","signature_b64":"3pUA4fW6B+CSErkuYQLXLQkDkWNbbs6XZcf8r/PonglYaNXtPp3c6CsJWm+tboH+ttOUDlCMPYhIezckjQnSDw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"8e103908e81925200be955163a8e06325f596760cbc7f03d1568f71c231e0d83","last_reissued_at":"2026-07-05T11:04:03.328268Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:04:03.328268Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"source_kind":"arxiv","source_id":"2505.10670","source_version":1,"attestation_state":"computed"},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-07-05T11:04:03Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"508EHHs0OEcSEcuupkmDkG/Nc6jxRRfFd9d4IWc+w+xltuKBfCUrMpp5YX+tTFoLihXxC9Hol7Yy8KXX78jBDQ==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-18T08:07:31.474674Z"},"content_sha256":"55cec3c4e309aabfba69a78ded09365d7ce569f0ff6a8abfcdb8eee91d3dd9ae","schema_version":"1.0","event_id":"sha256:55cec3c4e309aabfba69a78ded09365d7ce569f0ff6a8abfcdb8eee91d3dd9ae"},{"event_type":"graph_snapshot","subject_pith_number":"pith:2025:RYIDSCHIDESSAC7JKULDVDQGGJ","target":"graph","payload":{"graph_snapshot":{"paper":{"title":"Interpretable Risk Mitigation in LLM Agent Systems","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.CY","cs.GT"],"primary_cat":"cs.AI","authors_text":"Jan Chojnacki","submitted_at":"2025-05-15T19:22:11Z","abstract_excerpt":"Autonomous agents powered by large language models (LLMs) enable novel use cases in domains where responsible action is increasingly important. Yet the inherent unpredictability of LLMs raises safety concerns about agent reliability. In this work, we explore agent behaviour in a toy, game-theoretic environment based on a variation of the Iterated Prisoner's Dilemma. We introduce a strategy-modification method-independent of both the game and the prompt-by steering the residual stream with interpretable features extracted from a sparse autoencoder latent space. Steering with the good-faith nego"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2505.10670","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/2505.10670/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"},"verdict_id":null},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-07-05T11:04:03Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"veZ6e3jht1GP7Bcpx22NLvGnFLi8ZJdWBiMekcoO3ShxDWTRixCm03Bby3rjyfBj7Cb/uKzA/tVwv/y+KmDXBA==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-18T08:07:31.475182Z"},"content_sha256":"3a7eb955ec0378d8a1e5b0d4fd31b851e66edb7129c360bd3092acf17ef06a98","schema_version":"1.0","event_id":"sha256:3a7eb955ec0378d8a1e5b0d4fd31b851e66edb7129c360bd3092acf17ef06a98"}],"timestamp_proofs":[],"mirror_hints":[{"mirror_type":"https","name":"Pith Resolver","base_url":"https://pith.science","bundle_url":"https://pith.science/pith/RYIDSCHIDESSAC7JKULDVDQGGJ/bundle.json","state_url":"https://pith.science/pith/RYIDSCHIDESSAC7JKULDVDQGGJ/state.json","well_known_bundle_url":"https://pith.science/.well-known/pith/RYIDSCHIDESSAC7JKULDVDQGGJ/bundle.json","status":"primary"}],"public_keys":[{"key_id":"pith-v1-2026-05","algorithm":"ed25519","format":"raw","public_key_b64":"stVStoiQhXFxp4s2pdzPNoqVNBMojDU/fJ2db5S3CbM=","public_key_hex":"b2d552b68890857171a78b36a5dccf368a953413288c353f7c9d9d6f94b709b3","fingerprint_sha256_b32_first128bits":"RVFV5Z2OI2J3ZUO7ERDEBCYNKS","fingerprint_sha256_hex":"8d4b5ee74e4693bcd1df2446408b0d54","rotates_at":null,"url":"https://pith.science/pith-signing-key.json","notes":"Pith uses this Ed25519 key to sign canonical record SHA-256 digests. Verify with: ed25519_verify(public_key, message=canonical_sha256_bytes, signature=base64decode(signature_b64))."}],"merge_version":"pith-open-graph-merge-v1","built_at":"2026-08-18T08:07:31Z","links":{"resolver":"https://pith.science/pith/RYIDSCHIDESSAC7JKULDVDQGGJ","bundle":"https://pith.science/pith/RYIDSCHIDESSAC7JKULDVDQGGJ/bundle.json","state":"https://pith.science/pith/RYIDSCHIDESSAC7JKULDVDQGGJ/state.json","well_known_bundle":"https://pith.science/.well-known/pith/RYIDSCHIDESSAC7JKULDVDQGGJ/bundle.json"},"state":{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2025:RYIDSCHIDESSAC7JKULDVDQGGJ","merge_version":"pith-open-graph-merge-v1","event_count":2,"valid_event_count":2,"invalid_event_count":0,"equivocation_count":0,"current":{"canonical_record":{"metadata":{"abstract_canon_sha256":"4322dcdb486e874f3ed45032c9d4500e08c738e747dcd8c1ab312ee76cbdcbe3","cross_cats_sorted":["cs.CY","cs.GT"],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.AI","submitted_at":"2025-05-15T19:22:11Z","title_canon_sha256":"9f361c0122f74023b38c4c684377076ee708d55a27eb943a56315710d9e78a87"},"schema_version":"1.0","source":{"id":"2505.10670","kind":"arxiv","version":1}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2505.10670","created_at":"2026-07-05T11:04:03Z"},{"alias_kind":"arxiv_version","alias_value":"2505.10670v1","created_at":"2026-07-05T11:04:03Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2505.10670","created_at":"2026-07-05T11:04:03Z"},{"alias_kind":"pith_short_12","alias_value":"RYIDSCHIDESS","created_at":"2026-07-05T11:04:03Z"},{"alias_kind":"pith_short_16","alias_value":"RYIDSCHIDESSAC7J","created_at":"2026-07-05T11:04:03Z"},{"alias_kind":"pith_short_8","alias_value":"RYIDSCHI","created_at":"2026-07-05T11:04:03Z"}],"graph_snapshots":[{"event_id":"sha256:3a7eb955ec0378d8a1e5b0d4fd31b851e66edb7129c360bd3092acf17ef06a98","target":"graph","created_at":"2026-07-05T11:04:03Z","signer":{"key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signer_id":"pith.science","signer_type":"pith_registry"},"payload":{"graph_snapshot":{"author_claims":{"count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57","strong_count":0},"builder_version":"pith-number-builder-2026-05-17-v1","claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"formal_canon":{"evidence_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"integrity":{"available":true,"clean":true,"detectors_run":[],"endpoint":"/pith/2505.10670/integrity.json","findings":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"Autonomous agents powered by large language models (LLMs) enable novel use cases in domains where responsible action is increasingly important. Yet the inherent unpredictability of LLMs raises safety concerns about agent reliability. In this work, we explore agent behaviour in a toy, game-theoretic environment based on a variation of the Iterated Prisoner's Dilemma. We introduce a strategy-modification method-independent of both the game and the prompt-by steering the residual stream with interpretable features extracted from a sparse autoencoder latent space. Steering with the good-faith nego","authors_text":"Jan Chojnacki","cross_cats":["cs.CY","cs.GT"],"headline":"","license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.AI","submitted_at":"2025-05-15T19:22:11Z","title":"Interpretable Risk Mitigation in LLM Agent Systems"},"references":{"count":0,"internal_anchors":0,"resolved_work":0,"sample":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2505.10670","kind":"arxiv","version":1},"verdict":{"created_at":null,"id":null,"model_set":{},"one_line_summary":"","pipeline_version":null,"pith_extraction_headline":"","strongest_claim":"","weakest_assumption":""}},"verdict_id":null}}],"author_attestations":[],"timestamp_anchors":[],"storage_attestations":[],"citation_signatures":[],"replication_records":[],"corrections":[],"mirror_hints":[],"record_created":{"event_id":"sha256:55cec3c4e309aabfba69a78ded09365d7ce569f0ff6a8abfcdb8eee91d3dd9ae","target":"record","created_at":"2026-07-05T11:04:03Z","signer":{"key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signer_id":"pith.science","signer_type":"pith_registry"},"payload":{"attestation_state":"computed","canonical_record":{"metadata":{"abstract_canon_sha256":"4322dcdb486e874f3ed45032c9d4500e08c738e747dcd8c1ab312ee76cbdcbe3","cross_cats_sorted":["cs.CY","cs.GT"],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.AI","submitted_at":"2025-05-15T19:22:11Z","title_canon_sha256":"9f361c0122f74023b38c4c684377076ee708d55a27eb943a56315710d9e78a87"},"schema_version":"1.0","source":{"id":"2505.10670","kind":"arxiv","version":1}},"canonical_sha256":"8e103908e81925200be955163a8e06325f596760cbc7f03d1568f71c231e0d83","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"8e103908e81925200be955163a8e06325f596760cbc7f03d1568f71c231e0d83","first_computed_at":"2026-07-05T11:04:03.328268Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-07-05T11:04:03.328268Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"3pUA4fW6B+CSErkuYQLXLQkDkWNbbs6XZcf8r/PonglYaNXtPp3c6CsJWm+tboH+ttOUDlCMPYhIezckjQnSDw==","signature_status":"signed_v1","signed_at":"2026-07-05T11:04:03.328805Z","signed_message":"canonical_sha256_bytes"},"source_id":"2505.10670","source_kind":"arxiv","source_version":1}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:55cec3c4e309aabfba69a78ded09365d7ce569f0ff6a8abfcdb8eee91d3dd9ae","sha256:3a7eb955ec0378d8a1e5b0d4fd31b851e66edb7129c360bd3092acf17ef06a98"],"state_sha256":"72f17b91bef518743bf5bdb4308957bb64c68091ca654d08a4d85523247077e6"},"bundle_signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"7VtPwetO8ZDaw26Mk+xUt3pWQ9WXbOnHsdimtxkw4KKeyCrXzX0qXI9LdJt0eFox3s4r6ZUGia11J9/y3u5RBA==","signed_message":"bundle_sha256_bytes","signed_at":"2026-08-18T08:07:31.480734Z","bundle_sha256":"a54c8d7ee8c896ed9040ee4ba8db593c816fae4a4ef9072bf25f69335982cb09"}}