{"bundle_type":"pith_open_graph_bundle","bundle_version":"1.0","pith_number":"pith:2024:5NCY3SIYUOD7WZFZIGUVQRI7CW","short_pith_number":"pith:5NCY3SIY","canonical_record":{"source":{"id":"2406.02577","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2024-05-28T23:28:28Z","cross_cats_sorted":["cs.CR","cs.LG"],"title_canon_sha256":"fdbd4880ff8a24ae73258165a1e3ef3d4f6ff8f6a7a2fbaed8d92922a23e7058","abstract_canon_sha256":"83159705757b5eb05238e3c86d92ecdf9c68475b39bf84dbe49aa60d0efd8537"},"schema_version":"1.0"},"canonical_sha256":"eb458dc918a387fb64b941a958451f15ae889cfd9c7752e0f7a998ce93cc9e5d","source":{"kind":"arxiv","id":"2406.02577","version":1},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2406.02577","created_at":"2026-07-05T08:27:27Z"},{"alias_kind":"arxiv_version","alias_value":"2406.02577v1","created_at":"2026-07-05T08:27:27Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2406.02577","created_at":"2026-07-05T08:27:27Z"},{"alias_kind":"pith_short_12","alias_value":"5NCY3SIYUOD7","created_at":"2026-07-05T08:27:27Z"},{"alias_kind":"pith_short_16","alias_value":"5NCY3SIYUOD7WZFZ","created_at":"2026-07-05T08:27:27Z"},{"alias_kind":"pith_short_8","alias_value":"5NCY3SIY","created_at":"2026-07-05T08:27:27Z"}],"events":[{"event_type":"record_created","subject_pith_number":"pith:2024:5NCY3SIYUOD7WZFZIGUVQRI7CW","target":"record","payload":{"canonical_record":{"source":{"id":"2406.02577","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2024-05-28T23:28:28Z","cross_cats_sorted":["cs.CR","cs.LG"],"title_canon_sha256":"fdbd4880ff8a24ae73258165a1e3ef3d4f6ff8f6a7a2fbaed8d92922a23e7058","abstract_canon_sha256":"83159705757b5eb05238e3c86d92ecdf9c68475b39bf84dbe49aa60d0efd8537"},"schema_version":"1.0"},"canonical_sha256":"eb458dc918a387fb64b941a958451f15ae889cfd9c7752e0f7a998ce93cc9e5d","receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T08:27:27.166403Z","signature_b64":"ZFFbZvk9m8JjBv+0KMBmTVF1ubNiPrSTUHbAUQjIatGWNn7mWCvXsF8HpoMuxcbBMEBBfImFyzMBv0BXVO4lDQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"eb458dc918a387fb64b941a958451f15ae889cfd9c7752e0f7a998ce93cc9e5d","last_reissued_at":"2026-07-05T08:27:27.166015Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T08:27:27.166015Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"source_kind":"arxiv","source_id":"2406.02577","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":"2026-07-05T08:27:27Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"HQ+lAHJfQqQ7OQiFVEeFF/Bmja2W5HKjINx1bbAoTsHiC00PUJiA3hEwo2A3cfggpRElLEPavOwQeiYbftDGCw==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-15T22:12:24.090949Z"},"content_sha256":"0d0c2be62ce636bb7618e696831e010739616e665ccf47038d9a5a7b71a22841","schema_version":"1.0","event_id":"sha256:0d0c2be62ce636bb7618e696831e010739616e665ccf47038d9a5a7b71a22841"},{"event_type":"graph_snapshot","subject_pith_number":"pith:2024:5NCY3SIYUOD7WZFZIGUVQRI7CW","target":"graph","payload":{"graph_snapshot":{"paper":{"title":"Are PPO-ed Language Models Hackable?","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.CR","cs.LG"],"primary_cat":"cs.CL","authors_text":"David Getzen, Suraj Anand","submitted_at":"2024-05-28T23:28:28Z","abstract_excerpt":"Numerous algorithms have been proposed to $\\textit{align}$ language models to remove undesirable behaviors. However, the challenges associated with a very large state space and creating a proper reward function often result in various jailbreaks. Our paper aims to examine this effect of reward in the controlled setting of positive sentiment language generation. Instead of online training of a reward model based on human feedback, we employ a statically learned sentiment classifier. We also consider a setting where our model's weights and activations are exposed to an end-user after training. W"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2406.02577","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/2406.02577/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":"2026-07-05T08:27:27Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"GkgvoyD2AzVNqV4URwZO4KTDRx1w/vE1S7sER5Hy5KfthL/I5lHSu5kOKLdiadVct5jWtx47xSuYQoRmsFGRCA==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-15T22:12:24.091521Z"},"content_sha256":"105186d946caba55c303635ee79fe57cf369793ce7616cf4562428d66b11abdb","schema_version":"1.0","event_id":"sha256:105186d946caba55c303635ee79fe57cf369793ce7616cf4562428d66b11abdb"}],"timestamp_proofs":[],"mirror_hints":[{"mirror_type":"https","name":"Pith Resolver","base_url":"https://pith.science","bundle_url":"https://pith.science/pith/5NCY3SIYUOD7WZFZIGUVQRI7CW/bundle.json","state_url":"https://pith.science/pith/5NCY3SIYUOD7WZFZIGUVQRI7CW/state.json","well_known_bundle_url":"https://pith.science/.well-known/pith/5NCY3SIYUOD7WZFZIGUVQRI7CW/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":"2026-08-15T22:12:24Z","links":{"resolver":"https://pith.science/pith/5NCY3SIYUOD7WZFZIGUVQRI7CW","bundle":"https://pith.science/pith/5NCY3SIYUOD7WZFZIGUVQRI7CW/bundle.json","state":"https://pith.science/pith/5NCY3SIYUOD7WZFZIGUVQRI7CW/state.json","well_known_bundle":"https://pith.science/.well-known/pith/5NCY3SIYUOD7WZFZIGUVQRI7CW/bundle.json"},"state":{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2024:5NCY3SIYUOD7WZFZIGUVQRI7CW","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":"83159705757b5eb05238e3c86d92ecdf9c68475b39bf84dbe49aa60d0efd8537","cross_cats_sorted":["cs.CR","cs.LG"],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2024-05-28T23:28:28Z","title_canon_sha256":"fdbd4880ff8a24ae73258165a1e3ef3d4f6ff8f6a7a2fbaed8d92922a23e7058"},"schema_version":"1.0","source":{"id":"2406.02577","kind":"arxiv","version":1}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2406.02577","created_at":"2026-07-05T08:27:27Z"},{"alias_kind":"arxiv_version","alias_value":"2406.02577v1","created_at":"2026-07-05T08:27:27Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2406.02577","created_at":"2026-07-05T08:27:27Z"},{"alias_kind":"pith_short_12","alias_value":"5NCY3SIYUOD7","created_at":"2026-07-05T08:27:27Z"},{"alias_kind":"pith_short_16","alias_value":"5NCY3SIYUOD7WZFZ","created_at":"2026-07-05T08:27:27Z"},{"alias_kind":"pith_short_8","alias_value":"5NCY3SIY","created_at":"2026-07-05T08:27:27Z"}],"graph_snapshots":[{"event_id":"sha256:105186d946caba55c303635ee79fe57cf369793ce7616cf4562428d66b11abdb","target":"graph","created_at":"2026-07-05T08:27:27Z","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/2406.02577/integrity.json","findings":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"Numerous algorithms have been proposed to $\\textit{align}$ language models to remove undesirable behaviors. However, the challenges associated with a very large state space and creating a proper reward function often result in various jailbreaks. Our paper aims to examine this effect of reward in the controlled setting of positive sentiment language generation. Instead of online training of a reward model based on human feedback, we employ a statically learned sentiment classifier. We also consider a setting where our model's weights and activations are exposed to an end-user after training. W","authors_text":"David Getzen, Suraj Anand","cross_cats":["cs.CR","cs.LG"],"headline":"","license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2024-05-28T23:28:28Z","title":"Are PPO-ed Language Models Hackable?"},"references":{"count":0,"internal_anchors":0,"resolved_work":0,"sample":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2406.02577","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":"sha256:0d0c2be62ce636bb7618e696831e010739616e665ccf47038d9a5a7b71a22841","target":"record","created_at":"2026-07-05T08:27:27Z","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":"83159705757b5eb05238e3c86d92ecdf9c68475b39bf84dbe49aa60d0efd8537","cross_cats_sorted":["cs.CR","cs.LG"],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2024-05-28T23:28:28Z","title_canon_sha256":"fdbd4880ff8a24ae73258165a1e3ef3d4f6ff8f6a7a2fbaed8d92922a23e7058"},"schema_version":"1.0","source":{"id":"2406.02577","kind":"arxiv","version":1}},"canonical_sha256":"eb458dc918a387fb64b941a958451f15ae889cfd9c7752e0f7a998ce93cc9e5d","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"eb458dc918a387fb64b941a958451f15ae889cfd9c7752e0f7a998ce93cc9e5d","first_computed_at":"2026-07-05T08:27:27.166015Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-07-05T08:27:27.166015Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"ZFFbZvk9m8JjBv+0KMBmTVF1ubNiPrSTUHbAUQjIatGWNn7mWCvXsF8HpoMuxcbBMEBBfImFyzMBv0BXVO4lDQ==","signature_status":"signed_v1","signed_at":"2026-07-05T08:27:27.166403Z","signed_message":"canonical_sha256_bytes"},"source_id":"2406.02577","source_kind":"arxiv","source_version":1}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:0d0c2be62ce636bb7618e696831e010739616e665ccf47038d9a5a7b71a22841","sha256:105186d946caba55c303635ee79fe57cf369793ce7616cf4562428d66b11abdb"],"state_sha256":"87ee68a08100425e99e3a896480f62e40cd7622e7315b37604e9eabb650a2fd8"},"bundle_signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"jeK5w1YQi0j0Kg24n3kXQmKPPuo4FfRDrADPOWA6VjnX4nCB8du1TL0Wr26D/F59sI3qBG4G4e5PlvANfC/yCA==","signed_message":"bundle_sha256_bytes","signed_at":"2026-08-15T22:12:24.097093Z","bundle_sha256":"19f4ba164158b88b5f201ccd75f49cbf74a3b930adccccd4108ce2dd94eb27a4"}}