{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:KFRPDIYDSJEGHA4EJK4HV6EONW","short_pith_number":"pith:KFRPDIYD","schema_version":"1.0","canonical_sha256":"5162f1a30392486383844ab87af88e6d90d1eb26446a3e48dc465f2bcf124758","source":{"kind":"arxiv","id":"2407.14561","version":4},"attestation_state":"computed","paper":{"title":"NNsight and NDIF: Democratizing Access to Open-Weight Foundation Model Internals","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.LG","authors_text":"Aaron Mueller, Adam Belfki, Alexander R. Loftus, Arjun Guha, Arnab Sen Sharma, Byron C. Wallace, Caden Juang, Can Rager, Carla Brodley, David Bau, Dmitrii Troitskii, Eric Todd, Francesca Lucchetti, Jaden Fiotto-Kaufman, Jannik Brinkmann, Jonathan Bell, Koyena Pal, Michael Ripa, Nikhil Prakash, Samuel Marks","submitted_at":"2024-07-18T17:59:01Z","abstract_excerpt":"We introduce NNsight and NDIF, technologies that work in tandem to enable scientific study of the representations and computations learned by very large neural networks. NNsight is an open-source system that extends PyTorch to introduce deferred remote execution. The National Deep Inference Fabric (NDIF) is a scalable inference service that executes NNsight requests, allowing users to share GPU resources and pretrained models. These technologies are enabled by the Intervention Graph, an architecture developed to decouple experimental design from model runtime. Together, this framework provides"},"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":"2407.14561","kind":"arxiv","version":4},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2024-07-18T17:59:01Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"f46c4932df2537497dbf4b01d8f44da18aac48675a939846e0424f347fee63a5","abstract_canon_sha256":"a7ed2d7c72fb28548e6c8c7a1e27c76f7ebc099b23c0a07d088ab96dd398fd4d"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T10:42:36.386971Z","signature_b64":"6QJNJ0tdQZa8+VO+LL13tP8Ly4LsePpaDx3wzbGh9Lck/7XLTWyXmN4Qc7BimuoU2F+CcgXWkursHqhM4KcqBg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"5162f1a30392486383844ab87af88e6d90d1eb26446a3e48dc465f2bcf124758","last_reissued_at":"2026-07-05T10:42:36.386452Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T10:42:36.386452Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"NNsight and NDIF: Democratizing Access to Open-Weight Foundation Model Internals","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.LG","authors_text":"Aaron Mueller, Adam Belfki, Alexander R. Loftus, Arjun Guha, Arnab Sen Sharma, Byron C. Wallace, Caden Juang, Can Rager, Carla Brodley, David Bau, Dmitrii Troitskii, Eric Todd, Francesca Lucchetti, Jaden Fiotto-Kaufman, Jannik Brinkmann, Jonathan Bell, Koyena Pal, Michael Ripa, Nikhil Prakash, Samuel Marks","submitted_at":"2024-07-18T17:59:01Z","abstract_excerpt":"We introduce NNsight and NDIF, technologies that work in tandem to enable scientific study of the representations and computations learned by very large neural networks. NNsight is an open-source system that extends PyTorch to introduce deferred remote execution. The National Deep Inference Fabric (NDIF) is a scalable inference service that executes NNsight requests, allowing users to share GPU resources and pretrained models. These technologies are enabled by the Intervention Graph, an architecture developed to decouple experimental design from model runtime. Together, this framework provides"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2407.14561","kind":"arxiv","version":4},"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/2407.14561/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":"2407.14561","created_at":"2026-07-05T10:42:36.386514+00:00"},{"alias_kind":"arxiv_version","alias_value":"2407.14561v4","created_at":"2026-07-05T10:42:36.386514+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2407.14561","created_at":"2026-07-05T10:42:36.386514+00:00"},{"alias_kind":"pith_short_12","alias_value":"KFRPDIYDSJEG","created_at":"2026-07-05T10:42:36.386514+00:00"},{"alias_kind":"pith_short_16","alias_value":"KFRPDIYDSJEGHA4E","created_at":"2026-07-05T10:42:36.386514+00:00"},{"alias_kind":"pith_short_8","alias_value":"KFRPDIYD","created_at":"2026-07-05T10:42:36.386514+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":7,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.09936","citing_title":"One Lens, Many Worlds : A Capability-Typed Interface for World-Model Interpretability","ref_index":22,"is_internal_anchor":false},{"citing_arxiv_id":"2606.28273","citing_title":"Vision-Default, Prior-Override: Causal Mechanisms of Perception-Knowledge Conflict in Vision-Language Models","ref_index":3,"is_internal_anchor":false},{"citing_arxiv_id":"2606.00566","citing_title":"Same Payload, Different Channel: Measuring Trust Asymmetry in Tool-Using Language Models","ref_index":26,"is_internal_anchor":false},{"citing_arxiv_id":"2606.09707","citing_title":"BrainSurgery: Reproducible and Reliable Declarative Weight Manipulations for Model Editing and Upcycling","ref_index":15,"is_internal_anchor":false},{"citing_arxiv_id":"2605.11093","citing_title":"Enabling Performant and Flexible Model-Internal Observability for LLM Inference","ref_index":10,"is_internal_anchor":false},{"citing_arxiv_id":"2604.22166","citing_title":"Fine-Grained Analysis of Shared Syntactic Mechanisms in Language Models","ref_index":10,"is_internal_anchor":false},{"citing_arxiv_id":"2604.12426","citing_title":"Do Transformers Use their Depth Adaptively? Evidence from a Relational Reasoning Task","ref_index":7,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/KFRPDIYDSJEGHA4EJK4HV6EONW","json":"https://pith.science/pith/KFRPDIYDSJEGHA4EJK4HV6EONW.json","graph_json":"https://pith.science/api/pith-number/KFRPDIYDSJEGHA4EJK4HV6EONW/graph.json","events_json":"https://pith.science/api/pith-number/KFRPDIYDSJEGHA4EJK4HV6EONW/events.json","paper":"https://pith.science/paper/KFRPDIYD"},"agent_actions":{"view_html":"https://pith.science/pith/KFRPDIYDSJEGHA4EJK4HV6EONW","download_json":"https://pith.science/pith/KFRPDIYDSJEGHA4EJK4HV6EONW.json","view_paper":"https://pith.science/paper/KFRPDIYD","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2407.14561&json=true","fetch_graph":"https://pith.science/api/pith-number/KFRPDIYDSJEGHA4EJK4HV6EONW/graph.json","fetch_events":"https://pith.science/api/pith-number/KFRPDIYDSJEGHA4EJK4HV6EONW/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/KFRPDIYDSJEGHA4EJK4HV6EONW/action/timestamp_anchor","attest_storage":"https://pith.science/pith/KFRPDIYDSJEGHA4EJK4HV6EONW/action/storage_attestation","attest_author":"https://pith.science/pith/KFRPDIYDSJEGHA4EJK4HV6EONW/action/author_attestation","sign_citation":"https://pith.science/pith/KFRPDIYDSJEGHA4EJK4HV6EONW/action/citation_signature","submit_replication":"https://pith.science/pith/KFRPDIYDSJEGHA4EJK4HV6EONW/action/replication_record"}},"created_at":"2026-07-05T10:42:36.386514+00:00","updated_at":"2026-07-05T10:42:36.386514+00:00"}