{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2024:UPAT3JQCIUJVCYHS3ACGTXFFEP","merge_version":"pith-open-graph-merge-v1","event_count":2,"valid_event_count":2,"invalid_event_count":0,"equivocation_count":0,"current":{"canonical_record":{"metadata":{"abstract_canon_sha256":"f00bffffe26e5b1c282893880845be5db0e2652950fad224bec2c430385fd3da","cross_cats_sorted":["cs.AI","cs.CL"],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2024-08-09T16:06:42Z","title_canon_sha256":"a1c5ea48af15d6dd66bb04ab12f2de102010e702c9478725bb6d6659c541ae0b"},"schema_version":"1.0","source":{"id":"2408.05147","kind":"arxiv","version":2}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2408.05147","created_at":"2026-05-17T23:38:53Z"},{"alias_kind":"arxiv_version","alias_value":"2408.05147v2","created_at":"2026-05-17T23:38:53Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2408.05147","created_at":"2026-05-17T23:38:53Z"},{"alias_kind":"pith_short_12","alias_value":"UPAT3JQCIUJV","created_at":"2026-05-18T12:33:37Z"},{"alias_kind":"pith_short_16","alias_value":"UPAT3JQCIUJVCYHS","created_at":"2026-05-18T12:33:37Z"},{"alias_kind":"pith_short_8","alias_value":"UPAT3JQC","created_at":"2026-05-18T12:33:37Z"}],"graph_snapshots":[{"event_id":"sha256:86661e0e7b7fcf382b4a58ba86df7c9e3f99eacb5866ce3fbf457802ed4ed7b6","target":"graph","created_at":"2026-05-17T23:38:53Z","signer":{"key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signer_id":"pith.science","signer_type":"pith_registry"},"payload":{"graph_snapshot":{"author_claims":{"count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57","strong_count":0},"builder_version":"pith-number-builder-2026-05-17-v1","claims":{"count":4,"items":[{"attestation":"unclaimed","claim_id":"C1","kind":"strongest_claim","source":"verdict.strongest_claim","status":"machine_extracted","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."},{"attestation":"unclaimed","claim_id":"C2","kind":"weakest_assumption","source":"verdict.weakest_assumption","status":"machine_extracted","text":"That standard reconstruction and sparsity metrics are sufficient to establish that the released SAEs will be useful for downstream interpretability and safety research."},{"attestation":"unclaimed","claim_id":"C3","kind":"one_line_summary","source":"verdict.one_line_summary","status":"machine_extracted","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."},{"attestation":"unclaimed","claim_id":"C4","kind":"headline","source":"verdict.pith_extraction.headline","status":"machine_extracted","text":"Gemma Scope releases JumpReLU sparse autoencoders for every layer and sub-layer of the Gemma 2 2B and 9B models."}],"snapshot_sha256":"6d98d96415b23724b95c2689eac0d071078d615623a41c02a648f34ff01e02c2"},"formal_canon":{"evidence_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"paper":{"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","authors_text":"Anca Dragan, Arthur Conmy, J\\'anos Kram\\'ar, Lewis Smith, Neel Nanda, Nicolas Sonnerat, Rohin Shah, Senthooran Rajamanoharan, Tom Lieberum, Vikrant Varma","cross_cats":["cs.AI","cs.CL"],"headline":"Gemma Scope releases JumpReLU sparse autoencoders for every layer and sub-layer of the Gemma 2 2B and 9B models.","license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2024-08-09T16:06:42Z","title":"Gemma Scope: Open Sparse Autoencoders Everywhere All At Once on Gemma 2"},"references":{"count":14,"internal_anchors":4,"resolved_work":14,"sample":[{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":1,"title":"Towards Automated Circuit Discovery for Mechanistic Interpretability","work_id":"8407204b-a28a-4ba8-ae13-e7c04d3497c4","year":2024},{"cited_arxiv_id":"2312.11805","doi":"","is_internal_anchor":true,"ref_index":2,"title":"Gemini: A Family of Highly Capable Multimodal Models","work_id":"83f7c85b-3f11-450f-ac0c-64d9745220b2","year":2024},{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":3,"title":"Universal neurons in gpt2 language models","work_id":"eb5860a6-61cb-43e3-8b19-4ddbd580abe0","year":2024},{"cited_arxiv_id":"1412.6980","doi":"","is_internal_anchor":true,"ref_index":4,"title":"Adam: A Method for Stochastic Optimization","work_id":"1910796d-9b52-4683-bf5c-de9632c1028b","year":2023},{"cited_arxiv_id":"1301.3781","doi":"10.18653/v1/2023.blackboxnlp-1","is_internal_anchor":true,"ref_index":5,"title":"Efficient Estimation of Word Representations in Vector Space","work_id":"59edaa01-a696-45b3-9a08-5eae777a799e","year":2005}],"snapshot_sha256":"9b45b0ce42befd474e566c7d159c4fc587b6cbdc987f566f554ce66e80424dee"},"source":{"id":"2408.05147","kind":"arxiv","version":2},"verdict":{"created_at":"2026-05-15T05:43:33.135686Z","id":"ab479788-8b9f-4331-a394-fd184d075edf","model_set":{"reader":"grok-4.3"},"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","pith_extraction_headline":"Gemma Scope releases JumpReLU sparse autoencoders for every layer and sub-layer of the Gemma 2 2B and 9B models.","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.","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."}},"verdict_id":"ab479788-8b9f-4331-a394-fd184d075edf"}}],"author_attestations":[],"timestamp_anchors":[],"storage_attestations":[],"citation_signatures":[],"replication_records":[],"corrections":[],"mirror_hints":[],"record_created":{"event_id":"sha256:99d4c6b6d5aaa06f864ef6457b40e555203af3c73333e77ae57a190f38ed087e","target":"record","created_at":"2026-05-17T23:38:53Z","signer":{"key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signer_id":"pith.science","signer_type":"pith_registry"},"payload":{"attestation_state":"computed","canonical_record":{"metadata":{"abstract_canon_sha256":"f00bffffe26e5b1c282893880845be5db0e2652950fad224bec2c430385fd3da","cross_cats_sorted":["cs.AI","cs.CL"],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2024-08-09T16:06:42Z","title_canon_sha256":"a1c5ea48af15d6dd66bb04ab12f2de102010e702c9478725bb6d6659c541ae0b"},"schema_version":"1.0","source":{"id":"2408.05147","kind":"arxiv","version":2}},"canonical_sha256":"a3c13da60245135160f2d80469dca523f04f957f0dc80cc61c3134d1d39b30ff","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"a3c13da60245135160f2d80469dca523f04f957f0dc80cc61c3134d1d39b30ff","first_computed_at":"2026-05-17T23:38:53.348400Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-05-17T23:38:53.348400Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"WgdYp8JYb/njCWyFYZla0CTKHjToyzFnKQzO9+KevM2zzpGNI6LGVzyU/crurGtQ6T4K6hrwEIWQ9v1tIONdCg==","signature_status":"signed_v1","signed_at":"2026-05-17T23:38:53.348981Z","signed_message":"canonical_sha256_bytes"},"source_id":"2408.05147","source_kind":"arxiv","source_version":2}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:99d4c6b6d5aaa06f864ef6457b40e555203af3c73333e77ae57a190f38ed087e","sha256:86661e0e7b7fcf382b4a58ba86df7c9e3f99eacb5866ce3fbf457802ed4ed7b6"],"state_sha256":"450ce9444583b0f21b53acf654721d798789ec31e709cc2594f0d3ba9d32615c"}