{"bundle_type":"pith_open_graph_bundle","bundle_version":"1.0","pith_number":"pith:2026:WWJFZ7B4WPXRHL3TYTN55IPQVI","short_pith_number":"pith:WWJFZ7B4","canonical_record":{"source":{"id":"2606.03131","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2026-06-02T04:18:08Z","cross_cats_sorted":[],"title_canon_sha256":"16f2d180df6c6ad2c8f011e9ef988e49998e79195da8b9495235d2dcf0e880a2","abstract_canon_sha256":"2b67adc44d2a162c0a701679912227bedca19f8ef41c22dcd67fc7d0a2412cd1"},"schema_version":"1.0"},"canonical_sha256":"b5925cfc3cb3ef13af73c4dbdea1f0aa247adbc65b8410e6dea99766f8c113b3","source":{"kind":"arxiv","id":"2606.03131","version":1},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2606.03131","created_at":"2026-06-03T01:05:32Z"},{"alias_kind":"arxiv_version","alias_value":"2606.03131v1","created_at":"2026-06-03T01:05:32Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2606.03131","created_at":"2026-06-03T01:05:32Z"},{"alias_kind":"pith_short_12","alias_value":"WWJFZ7B4WPXR","created_at":"2026-06-03T01:05:32Z"},{"alias_kind":"pith_short_16","alias_value":"WWJFZ7B4WPXRHL3T","created_at":"2026-06-03T01:05:32Z"},{"alias_kind":"pith_short_8","alias_value":"WWJFZ7B4","created_at":"2026-06-03T01:05:32Z"}],"events":[{"event_type":"record_created","subject_pith_number":"pith:2026:WWJFZ7B4WPXRHL3TYTN55IPQVI","target":"record","payload":{"canonical_record":{"source":{"id":"2606.03131","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2026-06-02T04:18:08Z","cross_cats_sorted":[],"title_canon_sha256":"16f2d180df6c6ad2c8f011e9ef988e49998e79195da8b9495235d2dcf0e880a2","abstract_canon_sha256":"2b67adc44d2a162c0a701679912227bedca19f8ef41c22dcd67fc7d0a2412cd1"},"schema_version":"1.0"},"canonical_sha256":"b5925cfc3cb3ef13af73c4dbdea1f0aa247adbc65b8410e6dea99766f8c113b3","receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-06-03T01:05:32.538057Z","signature_b64":"z1W7sOXvCk4x4BxNBZdL+KVuzUep//A2urgS8xQ4tU/BBsE6WKfQ1zSd6Lixf7gIs0XXS/h4ZoLBfItCMhMHDg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"b5925cfc3cb3ef13af73c4dbdea1f0aa247adbc65b8410e6dea99766f8c113b3","last_reissued_at":"2026-06-03T01:05:32.537698Z","signature_status":"signed_v1","first_computed_at":"2026-06-03T01:05:32.537698Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"source_kind":"arxiv","source_id":"2606.03131","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":"2026-06-03T01:05:32Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"QQQQ0V3wUG+d/M7S+m1G0bu1zBXZTflfIIz2hSYEH6VqflXYuhDiX51hYSMmCyVqr88jdnst1BtcW1d05HsmCg==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-06-07T17:29:41.822221Z"},"content_sha256":"5bce97dbb1e6f60bc60ef075dc8c7db2d3fa7054aa81d971c35311b61f480615","schema_version":"1.0","event_id":"sha256:5bce97dbb1e6f60bc60ef075dc8c7db2d3fa7054aa81d971c35311b61f480615"},{"event_type":"graph_snapshot","subject_pith_number":"pith:2026:WWJFZ7B4WPXRHL3TYTN55IPQVI","target":"graph","payload":{"graph_snapshot":{"paper":{"title":"HARVE: Hacking-Aware Reward-Head Vector Editing for Robust Reward Models","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Caiyue Huang, Mengnan Du, Qiuyang Zhao, Shuang Liu, Xiaorong Chen, Yanguang Liu, Yuxuan Bo","submitted_at":"2026-06-02T04:18:08Z","abstract_excerpt":"Reward models are central to large language model (LLM) alignment, but they remain vulnerable to reward hacking. To evaluate reward-model robustness, we introduce RewardHackBench containing 13 reward-hacking patterns covering real life high-stakes domains and general settings, and we find severe failures on specific subcategories across eight reward models. To mitigate these failures, we propose HARVE, a training-free reward-head editing method for scalar reward models. Instead of fine-tuning the reward model, HARVE identifies a multi-directional hacking subspace from residual stream direction"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2606.03131","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/2606.03131/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":"2026-06-03T01:05:32Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"XHu7iWcNeshCkiEi6zDlyFS7VnWz84M4NEFpfJPNjB8qarOAKoIHSpZHdTCakjuS8y2T6M1IjUi2bDMAssPjBQ==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-06-07T17:29:41.822992Z"},"content_sha256":"324d92a3f87f72a56aef785526e36fc72496e54ee9cab8fe0f8eeeff2169ff87","schema_version":"1.0","event_id":"sha256:324d92a3f87f72a56aef785526e36fc72496e54ee9cab8fe0f8eeeff2169ff87"}],"timestamp_proofs":[],"mirror_hints":[{"mirror_type":"https","name":"Pith Resolver","base_url":"https://pith.science","bundle_url":"https://pith.science/pith/WWJFZ7B4WPXRHL3TYTN55IPQVI/bundle.json","state_url":"https://pith.science/pith/WWJFZ7B4WPXRHL3TYTN55IPQVI/state.json","well_known_bundle_url":"https://pith.science/.well-known/pith/WWJFZ7B4WPXRHL3TYTN55IPQVI/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":"2026-06-07T17:29:41Z","links":{"resolver":"https://pith.science/pith/WWJFZ7B4WPXRHL3TYTN55IPQVI","bundle":"https://pith.science/pith/WWJFZ7B4WPXRHL3TYTN55IPQVI/bundle.json","state":"https://pith.science/pith/WWJFZ7B4WPXRHL3TYTN55IPQVI/state.json","well_known_bundle":"https://pith.science/.well-known/pith/WWJFZ7B4WPXRHL3TYTN55IPQVI/bundle.json"},"state":{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2026:WWJFZ7B4WPXRHL3TYTN55IPQVI","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":"2b67adc44d2a162c0a701679912227bedca19f8ef41c22dcd67fc7d0a2412cd1","cross_cats_sorted":[],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2026-06-02T04:18:08Z","title_canon_sha256":"16f2d180df6c6ad2c8f011e9ef988e49998e79195da8b9495235d2dcf0e880a2"},"schema_version":"1.0","source":{"id":"2606.03131","kind":"arxiv","version":1}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2606.03131","created_at":"2026-06-03T01:05:32Z"},{"alias_kind":"arxiv_version","alias_value":"2606.03131v1","created_at":"2026-06-03T01:05:32Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2606.03131","created_at":"2026-06-03T01:05:32Z"},{"alias_kind":"pith_short_12","alias_value":"WWJFZ7B4WPXR","created_at":"2026-06-03T01:05:32Z"},{"alias_kind":"pith_short_16","alias_value":"WWJFZ7B4WPXRHL3T","created_at":"2026-06-03T01:05:32Z"},{"alias_kind":"pith_short_8","alias_value":"WWJFZ7B4","created_at":"2026-06-03T01:05:32Z"}],"graph_snapshots":[{"event_id":"sha256:324d92a3f87f72a56aef785526e36fc72496e54ee9cab8fe0f8eeeff2169ff87","target":"graph","created_at":"2026-06-03T01:05:32Z","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/2606.03131/integrity.json","findings":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"Reward models are central to large language model (LLM) alignment, but they remain vulnerable to reward hacking. To evaluate reward-model robustness, we introduce RewardHackBench containing 13 reward-hacking patterns covering real life high-stakes domains and general settings, and we find severe failures on specific subcategories across eight reward models. To mitigate these failures, we propose HARVE, a training-free reward-head editing method for scalar reward models. Instead of fine-tuning the reward model, HARVE identifies a multi-directional hacking subspace from residual stream direction","authors_text":"Caiyue Huang, Mengnan Du, Qiuyang Zhao, Shuang Liu, Xiaorong Chen, Yanguang Liu, Yuxuan Bo","cross_cats":[],"headline":"","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2026-06-02T04:18:08Z","title":"HARVE: Hacking-Aware Reward-Head Vector Editing for Robust Reward Models"},"references":{"count":0,"internal_anchors":0,"resolved_work":0,"sample":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2606.03131","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":"sha256:5bce97dbb1e6f60bc60ef075dc8c7db2d3fa7054aa81d971c35311b61f480615","target":"record","created_at":"2026-06-03T01:05:32Z","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":"2b67adc44d2a162c0a701679912227bedca19f8ef41c22dcd67fc7d0a2412cd1","cross_cats_sorted":[],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2026-06-02T04:18:08Z","title_canon_sha256":"16f2d180df6c6ad2c8f011e9ef988e49998e79195da8b9495235d2dcf0e880a2"},"schema_version":"1.0","source":{"id":"2606.03131","kind":"arxiv","version":1}},"canonical_sha256":"b5925cfc3cb3ef13af73c4dbdea1f0aa247adbc65b8410e6dea99766f8c113b3","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"b5925cfc3cb3ef13af73c4dbdea1f0aa247adbc65b8410e6dea99766f8c113b3","first_computed_at":"2026-06-03T01:05:32.537698Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-06-03T01:05:32.537698Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"z1W7sOXvCk4x4BxNBZdL+KVuzUep//A2urgS8xQ4tU/BBsE6WKfQ1zSd6Lixf7gIs0XXS/h4ZoLBfItCMhMHDg==","signature_status":"signed_v1","signed_at":"2026-06-03T01:05:32.538057Z","signed_message":"canonical_sha256_bytes"},"source_id":"2606.03131","source_kind":"arxiv","source_version":1}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:5bce97dbb1e6f60bc60ef075dc8c7db2d3fa7054aa81d971c35311b61f480615","sha256:324d92a3f87f72a56aef785526e36fc72496e54ee9cab8fe0f8eeeff2169ff87"],"state_sha256":"bcdaacd37ba4a9273ec2a0d9e6448195ec62768a1ddf2d8ee886ab62e420515b"},"bundle_signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"VYdS0b0FbJUAAKAUg317ILOlgGVWtwZ7p4JOGtBFRRLSApCCIrekTATSCeBlh3jsa9cClt/kOhXEwVRg5Mf7Bw==","signed_message":"bundle_sha256_bytes","signed_at":"2026-06-07T17:29:41.826862Z","bundle_sha256":"78c6ec482c0362b556ec195bdaff72338188c2d47a7a20274e20c7f348e17b6c"}}