{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:PSQOCR56IDO2XCIRW3QUT2XE4L","short_pith_number":"pith:PSQOCR56","schema_version":"1.0","canonical_sha256":"7ca0e147be40ddab8911b6e149eae4e2e07d550a64d93d9053db97b9054be7a5","source":{"kind":"arxiv","id":"2601.04710","version":2},"attestation_state":"computed","paper":{"title":"Steering the Noise: Turning Random Perturbations into Effective Descent for Memory-Efficient LLM Fine-Tuning","license":"http://creativecommons.org/licenses/by-nc-nd/4.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"cs.CL","authors_text":"Feihu Jin, Shipeng Cen, Ying Tan","submitted_at":"2026-01-08T08:27:15Z","abstract_excerpt":"Fine-tuning large language models (LLMs) achieves strong performance but is often limited by the memory overhead of backpropagation. Zeroth-order (ZO) optimization avoids this overhead by estimating gradients through forward passes alone, yet it typically converges slowly because random Gaussian perturbations yield high-variance gradient estimates in high-dimensional parameter spaces. In this paper, we propose a plug-and-play framework that turns random perturbations into more effective descent directions. The key idea is to draw a small pool of candidate perturbations, evaluate their loss val"},"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":"2601.04710","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by-nc-nd/4.0/","primary_cat":"cs.CL","submitted_at":"2026-01-08T08:27:15Z","cross_cats_sorted":["cs.LG"],"title_canon_sha256":"a6c79fe0a2fcd4aa0832b90620fc4b827c90328a1f228d22484b1557c5e4428d","abstract_canon_sha256":"301b1699904a1871ec5625ae9cd9b6f3a354bcefd6f8d89423291b3976d3c490"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-06-11T01:09:26.075604Z","signature_b64":"8NJ9dPtlJCeyEhKYyJSSdZYdCHy0BEc94ZKizV5tQIqupn9Z1VnZC5d1SvHg+qBUwHY6WwwyVEknfdAsx0HMCg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"7ca0e147be40ddab8911b6e149eae4e2e07d550a64d93d9053db97b9054be7a5","last_reissued_at":"2026-06-11T01:09:26.074595Z","signature_status":"signed_v1","first_computed_at":"2026-06-11T01:09:26.074595Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Steering the Noise: Turning Random Perturbations into Effective Descent for Memory-Efficient LLM Fine-Tuning","license":"http://creativecommons.org/licenses/by-nc-nd/4.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"cs.CL","authors_text":"Feihu Jin, Shipeng Cen, Ying Tan","submitted_at":"2026-01-08T08:27:15Z","abstract_excerpt":"Fine-tuning large language models (LLMs) achieves strong performance but is often limited by the memory overhead of backpropagation. Zeroth-order (ZO) optimization avoids this overhead by estimating gradients through forward passes alone, yet it typically converges slowly because random Gaussian perturbations yield high-variance gradient estimates in high-dimensional parameter spaces. In this paper, we propose a plug-and-play framework that turns random perturbations into more effective descent directions. The key idea is to draw a small pool of candidate perturbations, evaluate their loss val"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2601.04710","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/2601.04710/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":"2601.04710","created_at":"2026-06-11T01:09:26.074732+00:00"},{"alias_kind":"arxiv_version","alias_value":"2601.04710v2","created_at":"2026-06-11T01:09:26.074732+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2601.04710","created_at":"2026-06-11T01:09:26.074732+00:00"},{"alias_kind":"pith_short_12","alias_value":"PSQOCR56IDO2","created_at":"2026-06-11T01:09:26.074732+00:00"},{"alias_kind":"pith_short_16","alias_value":"PSQOCR56IDO2XCIR","created_at":"2026-06-11T01:09:26.074732+00:00"},{"alias_kind":"pith_short_8","alias_value":"PSQOCR56","created_at":"2026-06-11T01:09:26.074732+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":0,"internal_anchor_count":0,"sample":[]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/PSQOCR56IDO2XCIRW3QUT2XE4L","json":"https://pith.science/pith/PSQOCR56IDO2XCIRW3QUT2XE4L.json","graph_json":"https://pith.science/api/pith-number/PSQOCR56IDO2XCIRW3QUT2XE4L/graph.json","events_json":"https://pith.science/api/pith-number/PSQOCR56IDO2XCIRW3QUT2XE4L/events.json","paper":"https://pith.science/paper/PSQOCR56"},"agent_actions":{"view_html":"https://pith.science/pith/PSQOCR56IDO2XCIRW3QUT2XE4L","download_json":"https://pith.science/pith/PSQOCR56IDO2XCIRW3QUT2XE4L.json","view_paper":"https://pith.science/paper/PSQOCR56","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2601.04710&json=true","fetch_graph":"https://pith.science/api/pith-number/PSQOCR56IDO2XCIRW3QUT2XE4L/graph.json","fetch_events":"https://pith.science/api/pith-number/PSQOCR56IDO2XCIRW3QUT2XE4L/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/PSQOCR56IDO2XCIRW3QUT2XE4L/action/timestamp_anchor","attest_storage":"https://pith.science/pith/PSQOCR56IDO2XCIRW3QUT2XE4L/action/storage_attestation","attest_author":"https://pith.science/pith/PSQOCR56IDO2XCIRW3QUT2XE4L/action/author_attestation","sign_citation":"https://pith.science/pith/PSQOCR56IDO2XCIRW3QUT2XE4L/action/citation_signature","submit_replication":"https://pith.science/pith/PSQOCR56IDO2XCIRW3QUT2XE4L/action/replication_record"}},"created_at":"2026-06-11T01:09:26.074732+00:00","updated_at":"2026-06-11T01:09:26.074732+00:00"}