{"bundle_type":"pith_open_graph_bundle","bundle_version":"1.0","pith_number":"pith:2024:UVCT4IP6WU7VCPDAA6XXGQT472","short_pith_number":"pith:UVCT4IP6","canonical_record":{"source":{"id":"2410.14425","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2024-10-18T12:39:32Z","cross_cats_sorted":["cs.AI","cs.CR"],"title_canon_sha256":"276d66afb28ff2d645b8dae26ced02a6cce70be7835febfc4933e2e3f9cf2228","abstract_canon_sha256":"a17da0cdafdfaf4b39bf3532b2b51af85e2067d445d57f4bbb0c8727e6df8beb"},"schema_version":"1.0"},"canonical_sha256":"a5453e21feb53f513c6007af73427cfebcb1186d6e09bbe4459fa84dc4802eab","source":{"kind":"arxiv","id":"2410.14425","version":2},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2410.14425","created_at":"2026-07-05T11:05:35Z"},{"alias_kind":"arxiv_version","alias_value":"2410.14425v2","created_at":"2026-07-05T11:05:35Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2410.14425","created_at":"2026-07-05T11:05:35Z"},{"alias_kind":"pith_short_12","alias_value":"UVCT4IP6WU7V","created_at":"2026-07-05T11:05:35Z"},{"alias_kind":"pith_short_16","alias_value":"UVCT4IP6WU7VCPDA","created_at":"2026-07-05T11:05:35Z"},{"alias_kind":"pith_short_8","alias_value":"UVCT4IP6","created_at":"2026-07-05T11:05:35Z"}],"events":[{"event_type":"record_created","subject_pith_number":"pith:2024:UVCT4IP6WU7VCPDAA6XXGQT472","target":"record","payload":{"canonical_record":{"source":{"id":"2410.14425","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2024-10-18T12:39:32Z","cross_cats_sorted":["cs.AI","cs.CR"],"title_canon_sha256":"276d66afb28ff2d645b8dae26ced02a6cce70be7835febfc4933e2e3f9cf2228","abstract_canon_sha256":"a17da0cdafdfaf4b39bf3532b2b51af85e2067d445d57f4bbb0c8727e6df8beb"},"schema_version":"1.0"},"canonical_sha256":"a5453e21feb53f513c6007af73427cfebcb1186d6e09bbe4459fa84dc4802eab","receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:05:35.732669Z","signature_b64":"Dn7kAccaY8sFkIcg+6V+iiJlCny3X96vbn3r9cA4TiPY1lkooRRistnQ1HsdRvA7U12UlgafcLJpEBS5nOUbBQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"a5453e21feb53f513c6007af73427cfebcb1186d6e09bbe4459fa84dc4802eab","last_reissued_at":"2026-07-05T11:05:35.732074Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:05:35.732074Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"source_kind":"arxiv","source_id":"2410.14425","source_version":2,"attestation_state":"computed"},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-07-05T11:05:35Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"nw6pMjPCk7veAfs2iyc5aW8njnsGfbEJ8LiDCkNhFuUkB2/p0OqcvaBTN+zGFwJLF16SzS8P3pGCtuASPU1mCg==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-05T22:54:00.123876Z"},"content_sha256":"e239124aa5b839e621ac5076a864ec9e11355dad8ca27df82741001a3f004861","schema_version":"1.0","event_id":"sha256:e239124aa5b839e621ac5076a864ec9e11355dad8ca27df82741001a3f004861"},{"event_type":"graph_snapshot","subject_pith_number":"pith:2024:UVCT4IP6WU7VCPDAA6XXGQT472","target":"graph","payload":{"graph_snapshot":{"paper":{"title":"Unlearning Backdoor Attacks for LLMs with Weak-to-Strong Knowledge Distillation","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","cs.CR"],"primary_cat":"cs.CL","authors_text":"Cong-Duy Nguyen, Luu Anh Tuan, Meihuizi Jia, Shuai Zhao, Xiaobao Wu, Yanhao Jia, Yichao Feng","submitted_at":"2024-10-18T12:39:32Z","abstract_excerpt":"Parameter-efficient fine-tuning (PEFT) can bridge the gap between large language models (LLMs) and downstream tasks. However, PEFT has been proven vulnerable to malicious attacks. Research indicates that poisoned LLMs, even after PEFT, retain the capability to activate internalized backdoors when input samples contain predefined triggers. In this paper, we introduce a novel weak-to-strong unlearning algorithm to defend against backdoor attacks based on feature alignment knowledge distillation, named W2SDefense. Specifically, we first train a small-scale language model through full-parameter fi"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2410.14425","kind":"arxiv","version":2},"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/2410.14425/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":"2026-07-05T11:05:35Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"8fawwm+HzoOgdcVqu3Q4RGeoA7rCx447yTCU/sBL/K1uRKBmVV0eQ/xO8enMwtfJnX8op53V9J0+gvnLgRh6CA==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-05T22:54:00.124512Z"},"content_sha256":"e5f7825c1af087b92e113ee8631b9d7781dda6730f2c4969afe5b7ba03519325","schema_version":"1.0","event_id":"sha256:e5f7825c1af087b92e113ee8631b9d7781dda6730f2c4969afe5b7ba03519325"}],"timestamp_proofs":[],"mirror_hints":[{"mirror_type":"https","name":"Pith Resolver","base_url":"https://pith.science","bundle_url":"https://pith.science/pith/UVCT4IP6WU7VCPDAA6XXGQT472/bundle.json","state_url":"https://pith.science/pith/UVCT4IP6WU7VCPDAA6XXGQT472/state.json","well_known_bundle_url":"https://pith.science/.well-known/pith/UVCT4IP6WU7VCPDAA6XXGQT472/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":"2026-08-05T22:54:00Z","links":{"resolver":"https://pith.science/pith/UVCT4IP6WU7VCPDAA6XXGQT472","bundle":"https://pith.science/pith/UVCT4IP6WU7VCPDAA6XXGQT472/bundle.json","state":"https://pith.science/pith/UVCT4IP6WU7VCPDAA6XXGQT472/state.json","well_known_bundle":"https://pith.science/.well-known/pith/UVCT4IP6WU7VCPDAA6XXGQT472/bundle.json"},"state":{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2024:UVCT4IP6WU7VCPDAA6XXGQT472","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":"a17da0cdafdfaf4b39bf3532b2b51af85e2067d445d57f4bbb0c8727e6df8beb","cross_cats_sorted":["cs.AI","cs.CR"],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2024-10-18T12:39:32Z","title_canon_sha256":"276d66afb28ff2d645b8dae26ced02a6cce70be7835febfc4933e2e3f9cf2228"},"schema_version":"1.0","source":{"id":"2410.14425","kind":"arxiv","version":2}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2410.14425","created_at":"2026-07-05T11:05:35Z"},{"alias_kind":"arxiv_version","alias_value":"2410.14425v2","created_at":"2026-07-05T11:05:35Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2410.14425","created_at":"2026-07-05T11:05:35Z"},{"alias_kind":"pith_short_12","alias_value":"UVCT4IP6WU7V","created_at":"2026-07-05T11:05:35Z"},{"alias_kind":"pith_short_16","alias_value":"UVCT4IP6WU7VCPDA","created_at":"2026-07-05T11:05:35Z"},{"alias_kind":"pith_short_8","alias_value":"UVCT4IP6","created_at":"2026-07-05T11:05:35Z"}],"graph_snapshots":[{"event_id":"sha256:e5f7825c1af087b92e113ee8631b9d7781dda6730f2c4969afe5b7ba03519325","target":"graph","created_at":"2026-07-05T11:05:35Z","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/2410.14425/integrity.json","findings":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"Parameter-efficient fine-tuning (PEFT) can bridge the gap between large language models (LLMs) and downstream tasks. However, PEFT has been proven vulnerable to malicious attacks. Research indicates that poisoned LLMs, even after PEFT, retain the capability to activate internalized backdoors when input samples contain predefined triggers. In this paper, we introduce a novel weak-to-strong unlearning algorithm to defend against backdoor attacks based on feature alignment knowledge distillation, named W2SDefense. Specifically, we first train a small-scale language model through full-parameter fi","authors_text":"Cong-Duy Nguyen, Luu Anh Tuan, Meihuizi Jia, Shuai Zhao, Xiaobao Wu, Yanhao Jia, Yichao Feng","cross_cats":["cs.AI","cs.CR"],"headline":"","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2024-10-18T12:39:32Z","title":"Unlearning Backdoor Attacks for LLMs with Weak-to-Strong Knowledge Distillation"},"references":{"count":0,"internal_anchors":0,"resolved_work":0,"sample":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2410.14425","kind":"arxiv","version":2},"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:e239124aa5b839e621ac5076a864ec9e11355dad8ca27df82741001a3f004861","target":"record","created_at":"2026-07-05T11:05:35Z","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":"a17da0cdafdfaf4b39bf3532b2b51af85e2067d445d57f4bbb0c8727e6df8beb","cross_cats_sorted":["cs.AI","cs.CR"],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2024-10-18T12:39:32Z","title_canon_sha256":"276d66afb28ff2d645b8dae26ced02a6cce70be7835febfc4933e2e3f9cf2228"},"schema_version":"1.0","source":{"id":"2410.14425","kind":"arxiv","version":2}},"canonical_sha256":"a5453e21feb53f513c6007af73427cfebcb1186d6e09bbe4459fa84dc4802eab","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"a5453e21feb53f513c6007af73427cfebcb1186d6e09bbe4459fa84dc4802eab","first_computed_at":"2026-07-05T11:05:35.732074Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-07-05T11:05:35.732074Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"Dn7kAccaY8sFkIcg+6V+iiJlCny3X96vbn3r9cA4TiPY1lkooRRistnQ1HsdRvA7U12UlgafcLJpEBS5nOUbBQ==","signature_status":"signed_v1","signed_at":"2026-07-05T11:05:35.732669Z","signed_message":"canonical_sha256_bytes"},"source_id":"2410.14425","source_kind":"arxiv","source_version":2}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:e239124aa5b839e621ac5076a864ec9e11355dad8ca27df82741001a3f004861","sha256:e5f7825c1af087b92e113ee8631b9d7781dda6730f2c4969afe5b7ba03519325"],"state_sha256":"f7f88f998bebf6042a82941c9750b78cfddc9fb1f1863fb24123f4216334c37e"},"bundle_signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"OdE+ZAvnzmSnLAX9BYXUwWDvZJSukspvRXIhAvjcXrA/v1ZwKufnHwuMgJfiENOKHP/SnmD+sTHJDbBqMe60CA==","signed_message":"bundle_sha256_bytes","signed_at":"2026-08-05T22:54:00.130530Z","bundle_sha256":"db71fd4e51171b1ad30324b124ed71b782b365e8313fb7f6027ea34d26e822e3"}}