Step 2: tool results

get_claim

{
  "include": [
    "arguments",
    "subclaims",
    "related",
    "dependents"
  ],
  "claim_id": "759728cb-0411-4140-925b-d97d8022e57f"
}

{"claim":{"id":"759728cb-0411-4140-925b-d97d8022e57f","kind":"claim","canonical_form":"Sparse autoencoders rarely outperform simple baselines on practical downstream interpretability tasks.","claim_type":"empirical_derived","state":"active","decomposition_status":"complete","importance":0.6,"contestation":0.55,"domains":[],"tags":[{"slug":"mechanistic-interpretability","name":"Mechanistic Interpretability"}],"steward_state":"running","stewarded_at":"2026-11-14T20:40:00.622Z","canonical_direction_note":"This affirmative wording captures the source's critique that current sparse autoencoders lack compelling, broadly practical advantages over baselines.","created_at":"2026-11-13T00:58:17.755Z","updated_at":"2026-11-14T20:39:56.903Z","page_url":"https://minerval.ai/claims/759728cb-0411-4140-925b-d97d8022e57f"},"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":"2026-11-13T00:58:17.921Z","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":"2026-11-14T20:39:56.905Z","paid":true,"summary":"First assessment. The claim is general ('rarely', 'practical downstream tasks'), so say which tasks the evidence covers: probing, steering, unlearning, auditing, model diffing. Read the sources linked…"}],"requests":[],"requests_total":0,"subclaim_count":0,"instance_count":2,"source_map":null,"related":[],"dependents":[{"id":"cb2d56c9-fcce-47d3-b2b3-5127f10f3293","text":"Under what conditions, if any, do sparse-autoencoder and other dictionary-learning features outperform linear probes and difference-of-means steering on labeled language-model tasks?","claim_type":"empirical_derived","node_kind":"question","relation_type":"answers","reasoning":"The general null answer across detection and steering: no condition has been identified in which dictionary features reliably beat simple baselines on practical tasks. Widely stated after AxBench and the 2025 negative results.","importance":0.4,"assessment_status":null,"assessment_confidence":null,"assessment_credence":null,"bounty_micro_usd":null,"checked":null,"formal":false}],"dependents_total":1,"subclaims":[],"subclaims_assessed":0,"arguments":[],"unplaced_argument_passages":[{"instance_id":"9327c619-05ec-4478-a5dc-00d3f2da8ffd","source":{"id":"447e476f-1376-48e2-9679-30bc4fa0e96c","url":"https://www.lesswrong.com/posts/4uXCAJNuPKtKBsi28/negative-results-for-saes-on-downstream-tasks","title":"Negative Results for SAEs On Downstream Tasks and Deprioritising SAE Research (GDM Mech Interp Team Progress Update #2)"},"verbatim_text":"But there have been several other parallel projects in the literature such as Kantamneni et al., Farrell et al., Wu et al., that found negative results on other forms of probing, unlearning and steering, respectively.","context":"Aggregating evidence across studies.","stance":"advances","direction":"for"}],"unplaced_argument_passages_total":1}