{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:ZR7RAZK65FZOKO3M64WYD62IQE","short_pith_number":"pith:ZR7RAZK6","schema_version":"1.0","canonical_sha256":"cc7f10655ee972e53b6cf72d81fb488125c915e3fa97bfb39dd95650c73509c6","source":{"kind":"arxiv","id":"2410.16314","version":4},"attestation_state":"computed","paper":{"title":"Steering Large Language Models using Conceptors: Improving Addition-Based Activation Engineering","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"cs.NE","authors_text":"Joris Postmus, Steven Abreu","submitted_at":"2024-10-09T10:09:37Z","abstract_excerpt":"Large language models have transformed AI, yet reliably controlling their outputs remains a challenge. This paper explores activation engineering, where outputs of pre-trained LLMs are controlled by manipulating their activations at inference time. Unlike traditional methods using a single steering vector, we introduce conceptors - mathematical constructs that represent sets of activation vectors as ellipsoidal regions. Conceptors act as soft projection matrices and offer more precise control over complex activation patterns. Our experiments demonstrate that conceptors outperform traditional m"},"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":"2410.16314","kind":"arxiv","version":4},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.NE","submitted_at":"2024-10-09T10:09:37Z","cross_cats_sorted":["cs.LG"],"title_canon_sha256":"74c2dfcf3b04d137e940679019a91d3556a3116b157a24f119f471cdb60be43f","abstract_canon_sha256":"6305b6d3bc290a75b97d0574d494f5bf4207b468000a08e80d9eb7595446a27d"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:01:27.838888Z","signature_b64":"tqhsNXz2ubh9QzGqHDtiwONfD3A+9d9fNgJO1axlhPDTK7qeKQyVXM5hE08l60b3lFe7Gc1Q2aytDwigmd7fDw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"cc7f10655ee972e53b6cf72d81fb488125c915e3fa97bfb39dd95650c73509c6","last_reissued_at":"2026-07-05T11:01:27.838379Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:01:27.838379Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Steering Large Language Models using Conceptors: Improving Addition-Based Activation Engineering","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"cs.NE","authors_text":"Joris Postmus, Steven Abreu","submitted_at":"2024-10-09T10:09:37Z","abstract_excerpt":"Large language models have transformed AI, yet reliably controlling their outputs remains a challenge. This paper explores activation engineering, where outputs of pre-trained LLMs are controlled by manipulating their activations at inference time. Unlike traditional methods using a single steering vector, we introduce conceptors - mathematical constructs that represent sets of activation vectors as ellipsoidal regions. Conceptors act as soft projection matrices and offer more precise control over complex activation patterns. Our experiments demonstrate that conceptors outperform traditional m"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2410.16314","kind":"arxiv","version":4},"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/2410.16314/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":"2410.16314","created_at":"2026-07-05T11:01:27.838442+00:00"},{"alias_kind":"arxiv_version","alias_value":"2410.16314v4","created_at":"2026-07-05T11:01:27.838442+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2410.16314","created_at":"2026-07-05T11:01:27.838442+00:00"},{"alias_kind":"pith_short_12","alias_value":"ZR7RAZK65FZO","created_at":"2026-07-05T11:01:27.838442+00:00"},{"alias_kind":"pith_short_16","alias_value":"ZR7RAZK65FZOKO3M","created_at":"2026-07-05T11:01:27.838442+00:00"},{"alias_kind":"pith_short_8","alias_value":"ZR7RAZK6","created_at":"2026-07-05T11:01:27.838442+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":7,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.03093","citing_title":"Decomposing how prompting steers behavior","ref_index":7,"is_internal_anchor":false},{"citing_arxiv_id":"2606.05194","citing_title":"Temporal Preference Concepts and their Functions in a Large Language Model","ref_index":89,"is_internal_anchor":false},{"citing_arxiv_id":"2605.15688","citing_title":"$\\alpha$-TCAV: A Unified Framework for Testing with Concept Activation Vectors","ref_index":131,"is_internal_anchor":false},{"citing_arxiv_id":"2601.14004","citing_title":"Locate, Steer, and Improve: A Practical Survey of Actionable Mechanistic Interpretability in Large Language Models","ref_index":242,"is_internal_anchor":false},{"citing_arxiv_id":"2605.10664","citing_title":"Prompt-Activation Duality: Improving Activation Steering via Attention-Level Interventions","ref_index":12,"is_internal_anchor":false},{"citing_arxiv_id":"2605.10664","citing_title":"Prompt-Activation Duality: Improving Activation Steering via Attention-Level Interventions","ref_index":12,"is_internal_anchor":false},{"citing_arxiv_id":"2604.19149","citing_title":"How Do Answer Tokens Read Reasoning Traces? Self-Reading Patterns in Thinking LLMs for Quantitative Reasoning","ref_index":25,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/ZR7RAZK65FZOKO3M64WYD62IQE","json":"https://pith.science/pith/ZR7RAZK65FZOKO3M64WYD62IQE.json","graph_json":"https://pith.science/api/pith-number/ZR7RAZK65FZOKO3M64WYD62IQE/graph.json","events_json":"https://pith.science/api/pith-number/ZR7RAZK65FZOKO3M64WYD62IQE/events.json","paper":"https://pith.science/paper/ZR7RAZK6"},"agent_actions":{"view_html":"https://pith.science/pith/ZR7RAZK65FZOKO3M64WYD62IQE","download_json":"https://pith.science/pith/ZR7RAZK65FZOKO3M64WYD62IQE.json","view_paper":"https://pith.science/paper/ZR7RAZK6","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2410.16314&json=true","fetch_graph":"https://pith.science/api/pith-number/ZR7RAZK65FZOKO3M64WYD62IQE/graph.json","fetch_events":"https://pith.science/api/pith-number/ZR7RAZK65FZOKO3M64WYD62IQE/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/ZR7RAZK65FZOKO3M64WYD62IQE/action/timestamp_anchor","attest_storage":"https://pith.science/pith/ZR7RAZK65FZOKO3M64WYD62IQE/action/storage_attestation","attest_author":"https://pith.science/pith/ZR7RAZK65FZOKO3M64WYD62IQE/action/author_attestation","sign_citation":"https://pith.science/pith/ZR7RAZK65FZOKO3M64WYD62IQE/action/citation_signature","submit_replication":"https://pith.science/pith/ZR7RAZK65FZOKO3M64WYD62IQE/action/replication_record"}},"created_at":"2026-07-05T11:01:27.838442+00:00","updated_at":"2026-07-05T11:01:27.838442+00:00"}