{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:OP7QAYCZNWGAXPO2DQLP6N4DUP","short_pith_number":"pith:OP7QAYCZ","schema_version":"1.0","canonical_sha256":"73ff0060596d8c0bbdda1c16ff3783a3f4a254712d8ac3c4663ddb6f248ead7a","source":{"kind":"arxiv","id":"2411.07641","version":1},"attestation_state":"computed","paper":{"title":"Top-$n\\sigma$: Not All Logits Are You Need","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Chenxia Tang, Hongli Xu, Jianchun Liu, Liusheng Huang","submitted_at":"2024-11-12T08:46:43Z","abstract_excerpt":"Large language models (LLMs) typically employ greedy decoding or low-temperature sampling for reasoning tasks, reflecting a perceived trade-off between diversity and accuracy. We challenge this convention by introducing top-$n\\sigma$, a novel sampling method that operates directly on pre-softmax logits by leveraging a statistical threshold. Our key insight is that logits naturally separate into a Gaussian-distributed noisy region and a distinct informative region, enabling efficient token filtering without complex probability manipulations. Unlike existing methods (e.g., top-$p$, min-$p$) that"},"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":"2411.07641","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2024-11-12T08:46:43Z","cross_cats_sorted":[],"title_canon_sha256":"d0d2c1b1937187104ce4d1ea35a8bdbd4cbba2f45aa6203daa8ea2c4de8f6f9c","abstract_canon_sha256":"5db941b5a28337e981b6570321305c2c14b3dc8782f913292761018009ee8625"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T09:34:26.665068Z","signature_b64":"JX5kI910+mu0ZyHOYpUcMHCUfqviTwSgm1FAoPvz6u8008MZtcxNm69jmpbFX3SsyxvZO9jg0bAy+mAXD6BIDA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"73ff0060596d8c0bbdda1c16ff3783a3f4a254712d8ac3c4663ddb6f248ead7a","last_reissued_at":"2026-07-05T09:34:26.664571Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T09:34:26.664571Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Top-$n\\sigma$: Not All Logits Are You Need","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Chenxia Tang, Hongli Xu, Jianchun Liu, Liusheng Huang","submitted_at":"2024-11-12T08:46:43Z","abstract_excerpt":"Large language models (LLMs) typically employ greedy decoding or low-temperature sampling for reasoning tasks, reflecting a perceived trade-off between diversity and accuracy. We challenge this convention by introducing top-$n\\sigma$, a novel sampling method that operates directly on pre-softmax logits by leveraging a statistical threshold. Our key insight is that logits naturally separate into a Gaussian-distributed noisy region and a distinct informative region, enabling efficient token filtering without complex probability manipulations. Unlike existing methods (e.g., top-$p$, min-$p$) that"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2411.07641","kind":"arxiv","version":1},"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/2411.07641/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":"2411.07641","created_at":"2026-07-05T09:34:26.664646+00:00"},{"alias_kind":"arxiv_version","alias_value":"2411.07641v1","created_at":"2026-07-05T09:34:26.664646+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2411.07641","created_at":"2026-07-05T09:34:26.664646+00:00"},{"alias_kind":"pith_short_12","alias_value":"OP7QAYCZNWGA","created_at":"2026-07-05T09:34:26.664646+00:00"},{"alias_kind":"pith_short_16","alias_value":"OP7QAYCZNWGAXPO2","created_at":"2026-07-05T09:34:26.664646+00:00"},{"alias_kind":"pith_short_8","alias_value":"OP7QAYCZ","created_at":"2026-07-05T09:34:26.664646+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":3,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2605.27268","citing_title":"Lost in Sampling: Assessing Lexical Reachability in LLMs via the Word Coverage Score (WCS)","ref_index":17,"is_internal_anchor":false},{"citing_arxiv_id":"2605.17187","citing_title":"PluRule: A Benchmark for Moderating Pluralistic Communities on Social Media","ref_index":188,"is_internal_anchor":false},{"citing_arxiv_id":"2605.05777","citing_title":"Estimating the Black-box LLM Uncertainty with Distribution-Aligned Adversarial Distillation","ref_index":77,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/OP7QAYCZNWGAXPO2DQLP6N4DUP","json":"https://pith.science/pith/OP7QAYCZNWGAXPO2DQLP6N4DUP.json","graph_json":"https://pith.science/api/pith-number/OP7QAYCZNWGAXPO2DQLP6N4DUP/graph.json","events_json":"https://pith.science/api/pith-number/OP7QAYCZNWGAXPO2DQLP6N4DUP/events.json","paper":"https://pith.science/paper/OP7QAYCZ"},"agent_actions":{"view_html":"https://pith.science/pith/OP7QAYCZNWGAXPO2DQLP6N4DUP","download_json":"https://pith.science/pith/OP7QAYCZNWGAXPO2DQLP6N4DUP.json","view_paper":"https://pith.science/paper/OP7QAYCZ","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2411.07641&json=true","fetch_graph":"https://pith.science/api/pith-number/OP7QAYCZNWGAXPO2DQLP6N4DUP/graph.json","fetch_events":"https://pith.science/api/pith-number/OP7QAYCZNWGAXPO2DQLP6N4DUP/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/OP7QAYCZNWGAXPO2DQLP6N4DUP/action/timestamp_anchor","attest_storage":"https://pith.science/pith/OP7QAYCZNWGAXPO2DQLP6N4DUP/action/storage_attestation","attest_author":"https://pith.science/pith/OP7QAYCZNWGAXPO2DQLP6N4DUP/action/author_attestation","sign_citation":"https://pith.science/pith/OP7QAYCZNWGAXPO2DQLP6N4DUP/action/citation_signature","submit_replication":"https://pith.science/pith/OP7QAYCZNWGAXPO2DQLP6N4DUP/action/replication_record"}},"created_at":"2026-07-05T09:34:26.664646+00:00","updated_at":"2026-07-05T09:34:26.664646+00:00"}