{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:STEN6G7AZKCH7EK234A6SLVODP","short_pith_number":"pith:STEN6G7A","schema_version":"1.0","canonical_sha256":"94c8df1be0ca847f915adf01e92eae1bc4999519b231d21b764f203f06fb960d","source":{"kind":"arxiv","id":"2410.17245","version":1},"attestation_state":"computed","paper":{"title":"Towards Reliable Evaluation of Behavior Steering Interventions in LLMs","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.CL"],"primary_cat":"cs.AI","authors_text":"David Krueger, Ekdeep Singh Lubana, Itamar Pres, Laura Ruis","submitted_at":"2024-10-22T17:59:39Z","abstract_excerpt":"Representation engineering methods have recently shown promise for enabling efficient steering of model behavior. However, evaluation pipelines for these methods have primarily relied on subjective demonstrations, instead of quantitative, objective metrics. We aim to take a step towards addressing this issue by advocating for four properties missing from current evaluations: (i) contexts sufficiently similar to downstream tasks should be used for assessing intervention quality; (ii) model likelihoods should be accounted for; (iii) evaluations should allow for standardized comparisons across di"},"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":"2410.17245","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.AI","submitted_at":"2024-10-22T17:59:39Z","cross_cats_sorted":["cs.CL"],"title_canon_sha256":"275ee9c2aa74b5218a29e59c94f1bbc6f1b2be15a9fc66994e5ef373d07a3b04","abstract_canon_sha256":"6309da80bf6b6e2cb5777bce0a5541fa23e50dc0ba01713afea87dc74b3d674d"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T09:24:14.082007Z","signature_b64":"/mtFZwwcv9SZtoXiDzdsUfYiFGAN/Tjxn+lZxeWnt1E31DQ0joxKSO7l0/IL0SKwS+UH19qT44G7wStc61Z+Ag==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"94c8df1be0ca847f915adf01e92eae1bc4999519b231d21b764f203f06fb960d","last_reissued_at":"2026-07-05T09:24:14.081598Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T09:24:14.081598Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Towards Reliable Evaluation of Behavior Steering Interventions in LLMs","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.CL"],"primary_cat":"cs.AI","authors_text":"David Krueger, Ekdeep Singh Lubana, Itamar Pres, Laura Ruis","submitted_at":"2024-10-22T17:59:39Z","abstract_excerpt":"Representation engineering methods have recently shown promise for enabling efficient steering of model behavior. However, evaluation pipelines for these methods have primarily relied on subjective demonstrations, instead of quantitative, objective metrics. We aim to take a step towards addressing this issue by advocating for four properties missing from current evaluations: (i) contexts sufficiently similar to downstream tasks should be used for assessing intervention quality; (ii) model likelihoods should be accounted for; (iii) evaluations should allow for standardized comparisons across di"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2410.17245","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/2410.17245/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":"2410.17245","created_at":"2026-07-05T09:24:14.081655+00:00"},{"alias_kind":"arxiv_version","alias_value":"2410.17245v1","created_at":"2026-07-05T09:24:14.081655+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2410.17245","created_at":"2026-07-05T09:24:14.081655+00:00"},{"alias_kind":"pith_short_12","alias_value":"STEN6G7AZKCH","created_at":"2026-07-05T09:24:14.081655+00:00"},{"alias_kind":"pith_short_16","alias_value":"STEN6G7AZKCH7EK2","created_at":"2026-07-05T09:24:14.081655+00:00"},{"alias_kind":"pith_short_8","alias_value":"STEN6G7A","created_at":"2026-07-05T09:24:14.081655+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":4,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.26155","citing_title":"Detecting and Controlling Sycophancy with Cascading Linear Features","ref_index":15,"is_internal_anchor":false},{"citing_arxiv_id":"2606.07696","citing_title":"Adversarial Robustness of Activation Steering in Large Language Models","ref_index":32,"is_internal_anchor":false},{"citing_arxiv_id":"2605.12412","citing_title":"Stories in Space: In-Context Learning Trajectories in Conceptual Belief Space","ref_index":30,"is_internal_anchor":false},{"citing_arxiv_id":"2605.12412","citing_title":"Stories in Space: In-Context Learning Trajectories in Conceptual Belief Space","ref_index":93,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/STEN6G7AZKCH7EK234A6SLVODP","json":"https://pith.science/pith/STEN6G7AZKCH7EK234A6SLVODP.json","graph_json":"https://pith.science/api/pith-number/STEN6G7AZKCH7EK234A6SLVODP/graph.json","events_json":"https://pith.science/api/pith-number/STEN6G7AZKCH7EK234A6SLVODP/events.json","paper":"https://pith.science/paper/STEN6G7A"},"agent_actions":{"view_html":"https://pith.science/pith/STEN6G7AZKCH7EK234A6SLVODP","download_json":"https://pith.science/pith/STEN6G7AZKCH7EK234A6SLVODP.json","view_paper":"https://pith.science/paper/STEN6G7A","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2410.17245&json=true","fetch_graph":"https://pith.science/api/pith-number/STEN6G7AZKCH7EK234A6SLVODP/graph.json","fetch_events":"https://pith.science/api/pith-number/STEN6G7AZKCH7EK234A6SLVODP/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/STEN6G7AZKCH7EK234A6SLVODP/action/timestamp_anchor","attest_storage":"https://pith.science/pith/STEN6G7AZKCH7EK234A6SLVODP/action/storage_attestation","attest_author":"https://pith.science/pith/STEN6G7AZKCH7EK234A6SLVODP/action/author_attestation","sign_citation":"https://pith.science/pith/STEN6G7AZKCH7EK234A6SLVODP/action/citation_signature","submit_replication":"https://pith.science/pith/STEN6G7AZKCH7EK234A6SLVODP/action/replication_record"}},"created_at":"2026-07-05T09:24:14.081655+00:00","updated_at":"2026-07-05T09:24:14.081655+00:00"}