Step 2: tool results

get_claim

{
  "include": [
    "arguments",
    "subclaims",
    "dependents"
  ],
  "claim_id": "fc567967-00d4-456b-b196-d188db1187c4"
}

{"claim":{"id":"fc567967-00d4-456b-b196-d188db1187c4","kind":"claim","canonical_form":"Sparse autoencoders can reveal learned model knowledge when its application context differs semantically from training data.","claim_type":"empirical_derived","state":"active","decomposition_status":"complete","importance":0.65,"contestation":0.6,"domains":[],"tags":[{"slug":"mechanistic-interpretability","name":"Mechanistic Interpretability"},{"slug":"latent-knowledge-in-language-models","name":"Latent Knowledge in Language Models"},{"slug":"large-language-models","name":"Large language models"}],"steward_state":"running","stewarded_at":"2026-11-14T20:41:02.122Z","canonical_direction_note":"The affirmative form states the source's thesis that sparse autoencoders can expose learned knowledge across a semantic gap between training and application contexts.","created_at":"2026-11-13T01:11:41.567Z","updated_at":"2026-11-14T20:39:56.996Z","page_url":"https://minerval.ai/claims/fc567967-00d4-456b-b196-d188db1187c4"},"assessment":null,"coverage":{"examined":false,"on":[],"note":"No judgment of this node is in force. That is not a finding either way."},"quantity":null,"threshold_of":null,"question":null,"formalization":null,"formalization_pending":null,"formalization_history":[],"verification":null,"lean_checks":[],"lean_checks_total":0,"waiting_messages":[{"trigger":"structure_and_assess","from":"the system","sender_node_id":null,"sent_at":"2026-11-13T01:11:41.624Z","paid":false,"summary":"New claim onboarded. Give it the structure and assessment its importance warrants; a minor or settled claim may need no decomposition."},{"trigger":"mandate_plan","from":"a mandate's plan","sender_node_id":null,"sent_at":"2026-11-14T20:39:56.998Z","paid":true,"summary":"First assessment. Find the experiment or experiments behind this claim in the sources linked to it (the hidden-objective auditing work, arXiv 2503.10965, is likely among them) and establish: what know…"}],"requests":[],"requests_total":0,"subclaim_count":0,"instance_count":1,"source_map":null,"subclaims":[],"subclaims_assessed":0,"arguments":[],"unplaced_argument_passages":[{"instance_id":"13e362c2-7608-4071-a97f-f048f89da336","source":{"id":"bf2135ba-f5be-48c4-82eb-271c5bc3db71","url":"https://arxiv.org/pdf/2503.10965","title":"Auditing language models for hidden objectives"},"verbatim_text":"Notably, the semantic search baseline cannot be interpreted as revealing hidden knowledge, because it does not use access to the model being audited. Thus, the efficacy of semantic search in this setting undermines our conclusion that the SAE was revealing hidden knowledge. It is possible that, like semantic search, the SAE features were merely detecting surface-level semantic similarity, such as content about AI behaviors.","context":"The document first raises the objection against its initial interpretation","stance":"criticizes","direction":null},{"instance_id":"51e03596-97a3-4a05-9894-ffd521f7c8c5","source":{"id":"bf2135ba-f5be-48c4-82eb-271c5bc3db71","url":"https://arxiv.org/pdf/2503.10965","title":"Auditing language models for hidden objectives"},"verbatim_text":"However, inspecting active SAE features also works in settings where semantic search fails: investigating demonstrations of RM-sycophantic behaviors (e.g. transcripts in which the assistant generates a chocolate-containing recipe) or just the Assistant control token itself (see Sections 5.3.1 and 5.3.3). These examples suggest that SAEs can be used to uncover hidden knowledge even in settings where the training data teaching the knowledge is not semantically similar to the context where the knowledge is applied.","context":null,"stance":"advances","direction":"for"}],"unplaced_argument_passages_total":2,"dependents":[{"id":"6b0dcfde-1ad4-4608-83ac-8e5c1dc1dbf2","text":"When do sparse autoencoders and related dictionary-learning methods outperform other methods for discovering language-model properties?","claim_type":"empirical_derived","node_kind":"question","relation_type":"answers","reasoning":"The condition proposed by the original hidden-objective audit for when SAE inspection beats semantic search of training data: when the context in which the model applies what it learned does not resemble the data that taught it.","importance":0.45,"assessment_status":null,"assessment_confidence":null,"assessment_credence":null,"bounty_micro_usd":null,"checked":null,"formal":false}],"dependents_total":1}