{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:UPAT3JQCIUJVCYHS3ACGTXFFEP","short_pith_number":"pith:UPAT3JQC","schema_version":"1.0","canonical_sha256":"a3c13da60245135160f2d80469dca523f04f957f0dc80cc61c3134d1d39b30ff","source":{"kind":"arxiv","id":"2408.05147","version":2},"attestation_state":"computed","paper":{"title":"Gemma Scope: Open Sparse Autoencoders Everywhere All At Once on Gemma 2","license":"http://creativecommons.org/licenses/by/4.0/","headline":"Gemma Scope releases JumpReLU sparse autoencoders for every layer and sub-layer of the Gemma 2 2B and 9B models.","cross_cats":["cs.AI","cs.CL"],"primary_cat":"cs.LG","authors_text":"Anca Dragan, Arthur Conmy, J\\'anos Kram\\'ar, Lewis Smith, Neel Nanda, Nicolas Sonnerat, Rohin Shah, Senthooran Rajamanoharan, Tom Lieberum, Vikrant Varma","submitted_at":"2024-08-09T16:06:42Z","abstract_excerpt":"Sparse autoencoders (SAEs) are an unsupervised method for learning a sparse decomposition of a neural network's latent representations into seemingly interpretable features. Despite recent excitement about their potential, research applications outside of industry are limited by the high cost of training a comprehensive suite of SAEs. In this work, we introduce Gemma Scope, an open suite of JumpReLU SAEs trained on all layers and sub-layers of Gemma 2 2B and 9B and select layers of Gemma 2 27B base models. We primarily train SAEs on the Gemma 2 pre-trained models, but additionally release SAEs"},"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":true,"formal_links_present":false},"canonical_record":{"source":{"id":"2408.05147","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2024-08-09T16:06:42Z","cross_cats_sorted":["cs.AI","cs.CL"],"title_canon_sha256":"a1c5ea48af15d6dd66bb04ab12f2de102010e702c9478725bb6d6659c541ae0b","abstract_canon_sha256":"f00bffffe26e5b1c282893880845be5db0e2652950fad224bec2c430385fd3da"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-17T23:38:53.348981Z","signature_b64":"WgdYp8JYb/njCWyFYZla0CTKHjToyzFnKQzO9+KevM2zzpGNI6LGVzyU/crurGtQ6T4K6hrwEIWQ9v1tIONdCg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"a3c13da60245135160f2d80469dca523f04f957f0dc80cc61c3134d1d39b30ff","last_reissued_at":"2026-05-17T23:38:53.348400Z","signature_status":"signed_v1","first_computed_at":"2026-05-17T23:38:53.348400Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Gemma Scope: Open Sparse Autoencoders Everywhere All At Once on Gemma 2","license":"http://creativecommons.org/licenses/by/4.0/","headline":"Gemma Scope releases JumpReLU sparse autoencoders for every layer and sub-layer of the Gemma 2 2B and 9B models.","cross_cats":["cs.AI","cs.CL"],"primary_cat":"cs.LG","authors_text":"Anca Dragan, Arthur Conmy, J\\'anos Kram\\'ar, Lewis Smith, Neel Nanda, Nicolas Sonnerat, Rohin Shah, Senthooran Rajamanoharan, Tom Lieberum, Vikrant Varma","submitted_at":"2024-08-09T16:06:42Z","abstract_excerpt":"Sparse autoencoders (SAEs) are an unsupervised method for learning a sparse decomposition of a neural network's latent representations into seemingly interpretable features. Despite recent excitement about their potential, research applications outside of industry are limited by the high cost of training a comprehensive suite of SAEs. In this work, we introduce Gemma Scope, an open suite of JumpReLU SAEs trained on all layers and sub-layers of Gemma 2 2B and 9B and select layers of Gemma 2 27B base models. We primarily train SAEs on the Gemma 2 pre-trained models, but additionally release SAEs"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"We introduce Gemma Scope, an open suite of JumpReLU SAEs trained on all layers and sub-layers of Gemma 2 2B and 9B and select layers of Gemma 2 27B base models.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"That standard reconstruction and sparsity metrics are sufficient to establish that the released SAEs will be useful for downstream interpretability and safety research.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"Gemma Scope supplies trained sparse autoencoders for all layers of Gemma 2 2B and 9B plus select 27B layers, with public weights and benchmark scores.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"Gemma Scope releases JumpReLU sparse autoencoders for every layer and sub-layer of the Gemma 2 2B and 9B models.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"6d98d96415b23724b95c2689eac0d071078d615623a41c02a648f34ff01e02c2"},"source":{"id":"2408.05147","kind":"arxiv","version":2},"verdict":{"id":"ab479788-8b9f-4331-a394-fd184d075edf","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-15T05:43:33.135686Z","strongest_claim":"We introduce Gemma Scope, an open suite of JumpReLU SAEs trained on all layers and sub-layers of Gemma 2 2B and 9B and select layers of Gemma 2 27B base models.","one_line_summary":"Gemma Scope supplies trained sparse autoencoders for all layers of Gemma 2 2B and 9B plus select 27B layers, with public weights and benchmark scores.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"That standard reconstruction and sparsity metrics are sufficient to establish that the released SAEs will be useful for downstream interpretability and safety research.","pith_extraction_headline":"Gemma Scope releases JumpReLU sparse autoencoders for every layer and sub-layer of the Gemma 2 2B and 9B models."},"references":{"count":14,"sample":[{"doi":"","year":2024,"title":"Towards Automated Circuit Discovery for Mechanistic Interpretability","work_id":"8407204b-a28a-4ba8-ae13-e7c04d3497c4","ref_index":1,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2024,"title":"Gemini: A Family of Highly Capable Multimodal Models","work_id":"83f7c85b-3f11-450f-ac0c-64d9745220b2","ref_index":2,"cited_arxiv_id":"2312.11805","is_internal_anchor":true},{"doi":"","year":2024,"title":"Universal neurons in gpt2 language models","work_id":"eb5860a6-61cb-43e3-8b19-4ddbd580abe0","ref_index":3,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2023,"title":"Adam: A Method for Stochastic Optimization","work_id":"1910796d-9b52-4683-bf5c-de9632c1028b","ref_index":4,"cited_arxiv_id":"1412.6980","is_internal_anchor":true},{"doi":"10.18653/v1/2023.blackboxnlp-1","year":2005,"title":"Efficient Estimation of Word Representations in Vector Space","work_id":"59edaa01-a696-45b3-9a08-5eae777a799e","ref_index":5,"cited_arxiv_id":"1301.3781","is_internal_anchor":true}],"resolved_work":14,"snapshot_sha256":"9b45b0ce42befd474e566c7d159c4fc587b6cbdc987f566f554ce66e80424dee","internal_anchors":4},"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":"2408.05147","created_at":"2026-05-17T23:38:53.348484+00:00"},{"alias_kind":"arxiv_version","alias_value":"2408.05147v2","created_at":"2026-05-17T23:38:53.348484+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2408.05147","created_at":"2026-05-17T23:38:53.348484+00:00"},{"alias_kind":"pith_short_12","alias_value":"UPAT3JQCIUJV","created_at":"2026-05-18T12:33:37.589309+00:00"},{"alias_kind":"pith_short_16","alias_value":"UPAT3JQCIUJVCYHS","created_at":"2026-05-18T12:33:37.589309+00:00"},{"alias_kind":"pith_short_8","alias_value":"UPAT3JQC","created_at":"2026-05-18T12:33:37.589309+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":33,"internal_anchor_count":33,"sample":[{"citing_arxiv_id":"2605.13930","citing_title":"Mechanistic Interpretability of EEG Foundation Models via Sparse Autoencoders","ref_index":12,"is_internal_anchor":true},{"citing_arxiv_id":"2502.14888","citing_title":"Beyond Cross-Modal Alignment: Measuring and Leveraging Modality Gap in Vision-Language Models","ref_index":18,"is_internal_anchor":true},{"citing_arxiv_id":"2605.22719","citing_title":"Reading Task Failure Off the Activations: A Sparse-Feature Audit of GPT-2 Small on Indirect Object Identification","ref_index":6,"is_internal_anchor":true},{"citing_arxiv_id":"2605.10933","citing_title":"DECO: Sparse Mixture-of-Experts with Dense-Comparable Performance on End-Side Devices","ref_index":148,"is_internal_anchor":true},{"citing_arxiv_id":"2605.12770","citing_title":"WriteSAE: Sparse Autoencoders for Recurrent State","ref_index":26,"is_internal_anchor":true},{"citing_arxiv_id":"2605.17493","citing_title":"Beyond Linear Superposition: Discovering Climate Features in AI Weather Models with KAN-SAE","ref_index":10,"is_internal_anchor":true},{"citing_arxiv_id":"2605.13930","citing_title":"Mechanistic Interpretability of EEG Foundation Models via Sparse Autoencoders","ref_index":12,"is_internal_anchor":true},{"citing_arxiv_id":"2509.18127","citing_title":"Safe-SAIL: Towards a Fine-grained Safety Landscape of Large Language Models via Sparse Autoencoder Interpretation Framework","ref_index":24,"is_internal_anchor":true},{"citing_arxiv_id":"2510.18184","citing_title":"ActivationReasoning: Logical Reasoning in Latent Activation Spaces","ref_index":10,"is_internal_anchor":true},{"citing_arxiv_id":"2511.01680","citing_title":"Making Interpretable Discoveries from Unstructured Data: A High-Dimensional Multiple Hypothesis Testing Approach","ref_index":53,"is_internal_anchor":true},{"citing_arxiv_id":"2605.13930","citing_title":"Mechanistic Interpretability of EEG Foundation Models via Sparse Autoencoders","ref_index":12,"is_internal_anchor":true},{"citing_arxiv_id":"2605.12770","citing_title":"WriteSAE: Sparse Autoencoders for Recurrent State","ref_index":38,"is_internal_anchor":true},{"citing_arxiv_id":"2604.19765","citing_title":"Do Hallucination Neurons Generalize? Evidence from Cross-Domain Transfer in LLMs","ref_index":27,"is_internal_anchor":true},{"citing_arxiv_id":"2604.04946","citing_title":"Sparse Autoencoders as a Steering Basis for Phase Synchronization in Graph-Based CFD Surrogates","ref_index":5,"is_internal_anchor":true},{"citing_arxiv_id":"2605.12770","citing_title":"WriteSAE: Sparse Autoencoders for Recurrent State","ref_index":38,"is_internal_anchor":true},{"citing_arxiv_id":"2605.12874","citing_title":"Descriptive Collision in Sparse Autoencoder Auto-Interpretability: When One Explanation Describes Many Features","ref_index":17,"is_internal_anchor":true},{"citing_arxiv_id":"2604.03436","citing_title":"MetaSAEs: Joint Training with a Decomposability Penalty Produces More Atomic Sparse Autoencoder Latents","ref_index":11,"is_internal_anchor":true},{"citing_arxiv_id":"2605.10933","citing_title":"DECO: Sparse Mixture-of-Experts with Dense-Comparable Performance on End-Side Devices","ref_index":148,"is_internal_anchor":true},{"citing_arxiv_id":"2605.03160","citing_title":"Pairwise matrices for sparse autoencoders: single-feature inspection mislabels causal axes","ref_index":11,"is_internal_anchor":true},{"citing_arxiv_id":"2605.08611","citing_title":"The Echo Amplifies the Knowledge: Somatic Marker Analogues in Language Models via Emotion Vector Re-Injection","ref_index":12,"is_internal_anchor":true},{"citing_arxiv_id":"2605.07922","citing_title":"Tree SAE: Learning Hierarchical Feature Structures in Sparse Autoencoders","ref_index":14,"is_internal_anchor":true},{"citing_arxiv_id":"2605.10933","citing_title":"DECO: Sparse Mixture-of-Experts with Dense-Comparable Performance on End-Side Devices","ref_index":148,"is_internal_anchor":true},{"citing_arxiv_id":"2605.05443","citing_title":"SLAM: Structural Linguistic Activation Marking for Language Models","ref_index":13,"is_internal_anchor":true},{"citing_arxiv_id":"2605.05115","citing_title":"Manifold Steering Reveals the Shared Geometry of Neural Network Representation and Behavior","ref_index":203,"is_internal_anchor":true},{"citing_arxiv_id":"2605.01192","citing_title":"Linear-Readout Floors and Threshold Recovery in Computation in Superposition","ref_index":10,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/UPAT3JQCIUJVCYHS3ACGTXFFEP","json":"https://pith.science/pith/UPAT3JQCIUJVCYHS3ACGTXFFEP.json","graph_json":"https://pith.science/api/pith-number/UPAT3JQCIUJVCYHS3ACGTXFFEP/graph.json","events_json":"https://pith.science/api/pith-number/UPAT3JQCIUJVCYHS3ACGTXFFEP/events.json","paper":"https://pith.science/paper/UPAT3JQC"},"agent_actions":{"view_html":"https://pith.science/pith/UPAT3JQCIUJVCYHS3ACGTXFFEP","download_json":"https://pith.science/pith/UPAT3JQCIUJVCYHS3ACGTXFFEP.json","view_paper":"https://pith.science/paper/UPAT3JQC","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2408.05147&json=true","fetch_graph":"https://pith.science/api/pith-number/UPAT3JQCIUJVCYHS3ACGTXFFEP/graph.json","fetch_events":"https://pith.science/api/pith-number/UPAT3JQCIUJVCYHS3ACGTXFFEP/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/UPAT3JQCIUJVCYHS3ACGTXFFEP/action/timestamp_anchor","attest_storage":"https://pith.science/pith/UPAT3JQCIUJVCYHS3ACGTXFFEP/action/storage_attestation","attest_author":"https://pith.science/pith/UPAT3JQCIUJVCYHS3ACGTXFFEP/action/author_attestation","sign_citation":"https://pith.science/pith/UPAT3JQCIUJVCYHS3ACGTXFFEP/action/citation_signature","submit_replication":"https://pith.science/pith/UPAT3JQCIUJVCYHS3ACGTXFFEP/action/replication_record"}},"created_at":"2026-05-17T23:38:53.348484+00:00","updated_at":"2026-05-17T23:38:53.348484+00:00"}