{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:BU5CLHHPXHBAFNX74SOLRGBLTI","short_pith_number":"pith:BU5CLHHP","schema_version":"1.0","canonical_sha256":"0d3a259cefb9c202b6ffe49cb8982b9a18990ea5851b58c15b4141e0130fd95d","source":{"kind":"arxiv","id":"2410.12299","version":2},"attestation_state":"computed","paper":{"title":"Semantics-Adaptive Activation Intervention for LLMs via Dynamic Steering Vectors","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Jingyuan Yang, Wei Peng, Weixuan Wang","submitted_at":"2024-10-16T06:58:49Z","abstract_excerpt":"Large language models (LLMs) have achieved remarkable performance across many tasks, yet aligning them with desired behaviors remains challenging. Activation intervention has emerged as an effective and economical method to modify the behavior of LLMs. Despite considerable interest in this area, current intervention methods exclusively employ a fixed steering vector to modify model activations, lacking adaptability to diverse input semantics. To address this limitation, we propose Semantics-Adaptive Dynamic Intervention (SADI), a novel method that constructs a dynamic steering vector to interv"},"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":"2410.12299","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2024-10-16T06:58:49Z","cross_cats_sorted":[],"title_canon_sha256":"12d030fe51083375788ee7b8519219a4321ea515f0caf1ea29a410ddc6ad47dd","abstract_canon_sha256":"67459be6b4f5c0aae895cf8fa992a12a34e188cfc186c6f7e5662abe0447331f"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T10:19:29.619251Z","signature_b64":"KCUxZiYJfJKwA664TASwkK3hjFl0J2XVv9kmFMV2LsqLflLGxbC4QaUoZYuTzJJ5KXDKJsT19c0I4zB+UoCVDA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"0d3a259cefb9c202b6ffe49cb8982b9a18990ea5851b58c15b4141e0130fd95d","last_reissued_at":"2026-07-05T10:19:29.618808Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T10:19:29.618808Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Semantics-Adaptive Activation Intervention for LLMs via Dynamic Steering Vectors","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Jingyuan Yang, Wei Peng, Weixuan Wang","submitted_at":"2024-10-16T06:58:49Z","abstract_excerpt":"Large language models (LLMs) have achieved remarkable performance across many tasks, yet aligning them with desired behaviors remains challenging. Activation intervention has emerged as an effective and economical method to modify the behavior of LLMs. Despite considerable interest in this area, current intervention methods exclusively employ a fixed steering vector to modify model activations, lacking adaptability to diverse input semantics. To address this limitation, we propose Semantics-Adaptive Dynamic Intervention (SADI), a novel method that constructs a dynamic steering vector to interv"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2410.12299","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/2410.12299/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"},"aliases":[{"alias_kind":"arxiv","alias_value":"2410.12299","created_at":"2026-07-05T10:19:29.618866+00:00"},{"alias_kind":"arxiv_version","alias_value":"2410.12299v2","created_at":"2026-07-05T10:19:29.618866+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2410.12299","created_at":"2026-07-05T10:19:29.618866+00:00"},{"alias_kind":"pith_short_12","alias_value":"BU5CLHHPXHBA","created_at":"2026-07-05T10:19:29.618866+00:00"},{"alias_kind":"pith_short_16","alias_value":"BU5CLHHPXHBAFNX7","created_at":"2026-07-05T10:19:29.618866+00:00"},{"alias_kind":"pith_short_8","alias_value":"BU5CLHHP","created_at":"2026-07-05T10:19:29.618866+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":7,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2607.01433","citing_title":"CreativityNeuro: Steering Language Model Weights to Improve Divergent Thinking and Reduce Mode Collapse","ref_index":33,"is_internal_anchor":false},{"citing_arxiv_id":"2607.01774","citing_title":"Subliminal Clocks: Latent Time Modelling in Diffusion Language Models","ref_index":25,"is_internal_anchor":false},{"citing_arxiv_id":"2606.31257","citing_title":"Decodable Is Not Grounded: A Vision-Ablation Arbiter for VLM Spatial Reasoning","ref_index":40,"is_internal_anchor":false},{"citing_arxiv_id":"2605.28664","citing_title":"Activation Steering for Synthetic Data Generation: The Role of Diversity in Downstream Safety Detection","ref_index":53,"is_internal_anchor":false},{"citing_arxiv_id":"2606.00726","citing_title":"Latent Reward Steering: An Adaptive Inference-Time Framework that Implicitly Promotes Cognitive Behaviors in Reasoning LLMs","ref_index":54,"is_internal_anchor":false},{"citing_arxiv_id":"2605.05953","citing_title":"Hallucination as an Anomaly: Dynamic Intervention via Probabilistic Circuits","ref_index":26,"is_internal_anchor":false},{"citing_arxiv_id":"2605.05678","citing_title":"Chain of Risk: Safety Failures in Large Reasoning Models and Mitigation via Adaptive Multi-Principle Steering","ref_index":31,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/BU5CLHHPXHBAFNX74SOLRGBLTI","json":"https://pith.science/pith/BU5CLHHPXHBAFNX74SOLRGBLTI.json","graph_json":"https://pith.science/api/pith-number/BU5CLHHPXHBAFNX74SOLRGBLTI/graph.json","events_json":"https://pith.science/api/pith-number/BU5CLHHPXHBAFNX74SOLRGBLTI/events.json","paper":"https://pith.science/paper/BU5CLHHP"},"agent_actions":{"view_html":"https://pith.science/pith/BU5CLHHPXHBAFNX74SOLRGBLTI","download_json":"https://pith.science/pith/BU5CLHHPXHBAFNX74SOLRGBLTI.json","view_paper":"https://pith.science/paper/BU5CLHHP","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2410.12299&json=true","fetch_graph":"https://pith.science/api/pith-number/BU5CLHHPXHBAFNX74SOLRGBLTI/graph.json","fetch_events":"https://pith.science/api/pith-number/BU5CLHHPXHBAFNX74SOLRGBLTI/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/BU5CLHHPXHBAFNX74SOLRGBLTI/action/timestamp_anchor","attest_storage":"https://pith.science/pith/BU5CLHHPXHBAFNX74SOLRGBLTI/action/storage_attestation","attest_author":"https://pith.science/pith/BU5CLHHPXHBAFNX74SOLRGBLTI/action/author_attestation","sign_citation":"https://pith.science/pith/BU5CLHHPXHBAFNX74SOLRGBLTI/action/citation_signature","submit_replication":"https://pith.science/pith/BU5CLHHPXHBAFNX74SOLRGBLTI/action/replication_record"}},"created_at":"2026-07-05T10:19:29.618866+00:00","updated_at":"2026-07-05T10:19:29.618866+00:00"}