{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:73G2B76HD6M6EKZAOCE2F6OE3I","short_pith_number":"pith:73G2B76H","schema_version":"1.0","canonical_sha256":"fecda0ffc71f99e22b207089a2f9c4da0f750b354423ff1066108d7988f3cd6c","source":{"kind":"arxiv","id":"2412.18497","version":2},"attestation_state":"computed","paper":{"title":"Neuron-Level Differentiation of Memorization and Generalization in Large Language Models","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Cheng-Yu Lin, Ching-Yu Tsai, Da-Cheng Juan, Heng-Yi Liu, Keng-Te Liao, Ko-Wei Huang, Shou-De Lin, Tzu-Ling Cheng, Yi-Fu Fu, Yi-Ting Yang, Yu-Chieh Tu","submitted_at":"2024-12-24T15:28:56Z","abstract_excerpt":"We investigate how Large Language Models (LLMs) distinguish between memorization and generalization at the neuron level. Through carefully designed tasks, we identify distinct neuron subsets responsible for each behavior. Experiments on both a GPT-2 model trained from scratch and a pretrained LLaMA-3.2 model fine-tuned with LoRA show consistent neuron-level specialization. We further demonstrate that inference-time interventions on these neurons can steer the model's behavior toward memorization or generalization. To assess robustness, we evaluate intra-task and inter-task consistency, confirm"},"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":"2412.18497","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2024-12-24T15:28:56Z","cross_cats_sorted":[],"title_canon_sha256":"6b3125e3f81987559b2451608360c0ccd7500fb3e530ac60d5cfb0f5db22a48b","abstract_canon_sha256":"ce40b85d35c84a4fb45a683fea48c9860e31f73627211ea5dd22bb621c291877"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:34:03.307412Z","signature_b64":"G2YCVYlYJPd1PVxMjAsropH3xtisbhE/wcaw7FtXt4OwyhuTSPA9qkqLpT6KWwZOV9K2t6IexpQCOe5U54cFDQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"fecda0ffc71f99e22b207089a2f9c4da0f750b354423ff1066108d7988f3cd6c","last_reissued_at":"2026-07-05T11:34:03.306909Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:34:03.306909Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Neuron-Level Differentiation of Memorization and Generalization in Large Language Models","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Cheng-Yu Lin, Ching-Yu Tsai, Da-Cheng Juan, Heng-Yi Liu, Keng-Te Liao, Ko-Wei Huang, Shou-De Lin, Tzu-Ling Cheng, Yi-Fu Fu, Yi-Ting Yang, Yu-Chieh Tu","submitted_at":"2024-12-24T15:28:56Z","abstract_excerpt":"We investigate how Large Language Models (LLMs) distinguish between memorization and generalization at the neuron level. Through carefully designed tasks, we identify distinct neuron subsets responsible for each behavior. Experiments on both a GPT-2 model trained from scratch and a pretrained LLaMA-3.2 model fine-tuned with LoRA show consistent neuron-level specialization. We further demonstrate that inference-time interventions on these neurons can steer the model's behavior toward memorization or generalization. To assess robustness, we evaluate intra-task and inter-task consistency, confirm"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2412.18497","kind":"arxiv","version":2},"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/2412.18497/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":"2412.18497","created_at":"2026-07-05T11:34:03.306973+00:00"},{"alias_kind":"arxiv_version","alias_value":"2412.18497v2","created_at":"2026-07-05T11:34:03.306973+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2412.18497","created_at":"2026-07-05T11:34:03.306973+00:00"},{"alias_kind":"pith_short_12","alias_value":"73G2B76HD6M6","created_at":"2026-07-05T11:34:03.306973+00:00"},{"alias_kind":"pith_short_16","alias_value":"73G2B76HD6M6EKZA","created_at":"2026-07-05T11:34:03.306973+00:00"},{"alias_kind":"pith_short_8","alias_value":"73G2B76H","created_at":"2026-07-05T11:34:03.306973+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2507.14777","citing_title":"Rethinking Memorization Measures and their Implications in Large Language Models","ref_index":61,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/73G2B76HD6M6EKZAOCE2F6OE3I","json":"https://pith.science/pith/73G2B76HD6M6EKZAOCE2F6OE3I.json","graph_json":"https://pith.science/api/pith-number/73G2B76HD6M6EKZAOCE2F6OE3I/graph.json","events_json":"https://pith.science/api/pith-number/73G2B76HD6M6EKZAOCE2F6OE3I/events.json","paper":"https://pith.science/paper/73G2B76H"},"agent_actions":{"view_html":"https://pith.science/pith/73G2B76HD6M6EKZAOCE2F6OE3I","download_json":"https://pith.science/pith/73G2B76HD6M6EKZAOCE2F6OE3I.json","view_paper":"https://pith.science/paper/73G2B76H","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2412.18497&json=true","fetch_graph":"https://pith.science/api/pith-number/73G2B76HD6M6EKZAOCE2F6OE3I/graph.json","fetch_events":"https://pith.science/api/pith-number/73G2B76HD6M6EKZAOCE2F6OE3I/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/73G2B76HD6M6EKZAOCE2F6OE3I/action/timestamp_anchor","attest_storage":"https://pith.science/pith/73G2B76HD6M6EKZAOCE2F6OE3I/action/storage_attestation","attest_author":"https://pith.science/pith/73G2B76HD6M6EKZAOCE2F6OE3I/action/author_attestation","sign_citation":"https://pith.science/pith/73G2B76HD6M6EKZAOCE2F6OE3I/action/citation_signature","submit_replication":"https://pith.science/pith/73G2B76HD6M6EKZAOCE2F6OE3I/action/replication_record"}},"created_at":"2026-07-05T11:34:03.306973+00:00","updated_at":"2026-07-05T11:34:03.306973+00:00"}