{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:LMYV4P2H3VYCWGVB5RU273CV6C","short_pith_number":"pith:LMYV4P2H","schema_version":"1.0","canonical_sha256":"5b315e3f47dd702b1aa1ec69afec55f0ab6b88c36029c58f4bd2b815c129aea5","source":{"kind":"arxiv","id":"2401.06102","version":4},"attestation_state":"computed","paper":{"title":"Patchscopes: A Unifying Framework for Inspecting Hidden Representations of Language Models","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.LG"],"primary_cat":"cs.CL","authors_text":"Adam Pearce, Asma Ghandeharioun, Avi Caciularu, Lucas Dixon, Mor Geva","submitted_at":"2024-01-11T18:33:48Z","abstract_excerpt":"Understanding the internal representations of large language models (LLMs) can help explain models' behavior and verify their alignment with human values. Given the capabilities of LLMs in generating human-understandable text, we propose leveraging the model itself to explain its internal representations in natural language. We introduce a framework called Patchscopes and show how it can be used to answer a wide range of questions about an LLM's computation. We show that many prior interpretability methods based on projecting representations into the vocabulary space and intervening on the LLM"},"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":"2401.06102","kind":"arxiv","version":4},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2024-01-11T18:33:48Z","cross_cats_sorted":["cs.AI","cs.LG"],"title_canon_sha256":"aa7f10ca25dc67537e51e885b31bd35b104dce99b93dadb90621c58c99a572db","abstract_canon_sha256":"f52885caa7c152cea14b1e95eba941b2ee30c901ee3d3d34142f52a3d4cdb461"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T08:28:36.882852Z","signature_b64":"1yNEmZPeNN8rkzTfPlxWozMBT99Ol2sKhvNlJkw4Zuj04oOwoem6t2hEETqsfoLGHLHyUKW9YcXy37wGDW6+BQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"5b315e3f47dd702b1aa1ec69afec55f0ab6b88c36029c58f4bd2b815c129aea5","last_reissued_at":"2026-07-05T08:28:36.882411Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T08:28:36.882411Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Patchscopes: A Unifying Framework for Inspecting Hidden Representations of Language Models","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.LG"],"primary_cat":"cs.CL","authors_text":"Adam Pearce, Asma Ghandeharioun, Avi Caciularu, Lucas Dixon, Mor Geva","submitted_at":"2024-01-11T18:33:48Z","abstract_excerpt":"Understanding the internal representations of large language models (LLMs) can help explain models' behavior and verify their alignment with human values. Given the capabilities of LLMs in generating human-understandable text, we propose leveraging the model itself to explain its internal representations in natural language. We introduce a framework called Patchscopes and show how it can be used to answer a wide range of questions about an LLM's computation. We show that many prior interpretability methods based on projecting representations into the vocabulary space and intervening on the LLM"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2401.06102","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/2401.06102/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":"2401.06102","created_at":"2026-07-05T08:28:36.882472+00:00"},{"alias_kind":"arxiv_version","alias_value":"2401.06102v4","created_at":"2026-07-05T08:28:36.882472+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2401.06102","created_at":"2026-07-05T08:28:36.882472+00:00"},{"alias_kind":"pith_short_12","alias_value":"LMYV4P2H3VYC","created_at":"2026-07-05T08:28:36.882472+00:00"},{"alias_kind":"pith_short_16","alias_value":"LMYV4P2H3VYCWGVB","created_at":"2026-07-05T08:28:36.882472+00:00"},{"alias_kind":"pith_short_8","alias_value":"LMYV4P2H","created_at":"2026-07-05T08:28:36.882472+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":12,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.23404","citing_title":"ReasoningLens: Hierarchical Visualization and Diagnostic Auditing for Large Reasoning Models","ref_index":2,"is_internal_anchor":false},{"citing_arxiv_id":"2606.21345","citing_title":"Factual Retrieval in LLMs Is a Redundant, Distributed and Non-Contiguous Process","ref_index":11,"is_internal_anchor":false},{"citing_arxiv_id":"2607.01571","citing_title":"Geometric Signatures of Reasoning: A Spectral Perspective on Task Hardness","ref_index":7,"is_internal_anchor":false},{"citing_arxiv_id":"2607.01033","citing_title":"The Model Organism Lottery: Model Organism Interpretability Strongly Depends on Training Methodology","ref_index":7,"is_internal_anchor":false},{"citing_arxiv_id":"2606.02780","citing_title":"Do Value Vectors in Deep Layers Need Context from the Residual Stream?","ref_index":74,"is_internal_anchor":false},{"citing_arxiv_id":"2605.24957","citing_title":"Mitigating Object Hallucinations in Vision-Language Models through Region-Aware Attention Recalibration","ref_index":12,"is_internal_anchor":false},{"citing_arxiv_id":"2606.09563","citing_title":"PRISM: Recovering Instruction Sets from Language Model Activations","ref_index":18,"is_internal_anchor":false},{"citing_arxiv_id":"2605.22462","citing_title":"From Correlation to Cause: A Five-Stage Methodology for Feature Analysis in Transformer Language Models","ref_index":8,"is_internal_anchor":false},{"citing_arxiv_id":"2509.13316","citing_title":"Do Activation Verbalization Methods Convey Privileged Information?","ref_index":18,"is_internal_anchor":false},{"citing_arxiv_id":"2510.01685","citing_title":"How Do Language Models Compose Functions?","ref_index":12,"is_internal_anchor":false},{"citing_arxiv_id":"2604.06377","citing_title":"The Master Key Hypothesis: Unlocking Cross-Model Capability Transfer via Linear Subspace Alignment","ref_index":20,"is_internal_anchor":false},{"citing_arxiv_id":"2605.01381","citing_title":"A framework for analyzing concept representations in neural models","ref_index":184,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/LMYV4P2H3VYCWGVB5RU273CV6C","json":"https://pith.science/pith/LMYV4P2H3VYCWGVB5RU273CV6C.json","graph_json":"https://pith.science/api/pith-number/LMYV4P2H3VYCWGVB5RU273CV6C/graph.json","events_json":"https://pith.science/api/pith-number/LMYV4P2H3VYCWGVB5RU273CV6C/events.json","paper":"https://pith.science/paper/LMYV4P2H"},"agent_actions":{"view_html":"https://pith.science/pith/LMYV4P2H3VYCWGVB5RU273CV6C","download_json":"https://pith.science/pith/LMYV4P2H3VYCWGVB5RU273CV6C.json","view_paper":"https://pith.science/paper/LMYV4P2H","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2401.06102&json=true","fetch_graph":"https://pith.science/api/pith-number/LMYV4P2H3VYCWGVB5RU273CV6C/graph.json","fetch_events":"https://pith.science/api/pith-number/LMYV4P2H3VYCWGVB5RU273CV6C/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/LMYV4P2H3VYCWGVB5RU273CV6C/action/timestamp_anchor","attest_storage":"https://pith.science/pith/LMYV4P2H3VYCWGVB5RU273CV6C/action/storage_attestation","attest_author":"https://pith.science/pith/LMYV4P2H3VYCWGVB5RU273CV6C/action/author_attestation","sign_citation":"https://pith.science/pith/LMYV4P2H3VYCWGVB5RU273CV6C/action/citation_signature","submit_replication":"https://pith.science/pith/LMYV4P2H3VYCWGVB5RU273CV6C/action/replication_record"}},"created_at":"2026-07-05T08:28:36.882472+00:00","updated_at":"2026-07-05T08:28:36.882472+00:00"}