{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:EVQ4XBCU23LVFG4ID5CFA4PTZW","short_pith_number":"pith:EVQ4XBCU","schema_version":"1.0","canonical_sha256":"2561cb8454d6d7529b881f445071f3cdb92f14bf8dfe3ff5fc8f2ab89a04c899","source":{"kind":"arxiv","id":"2511.11041","version":2},"attestation_state":"computed","paper":{"title":"Correcting Mean Bias in Text Embeddings: A Refined Renormalization with Training-Free Improvements on MMTEB","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","cs.LG"],"primary_cat":"cs.CL","authors_text":"Haoyu Liang, Xingyu Ren, Youran Sun","submitted_at":"2025-11-14T07:51:59Z","abstract_excerpt":"We find that current sentence-embedding models produce outputs with a consistent bias: every embedding $e$ decomposes as $\\tilde e + \\mu$, where the mean $\\mu$ is near-identical across all sentences. We study two training-free corrections -- subtracting $\\mu$ directly (R1), or projecting each embedding off the mean direction (R2) -- and show, via a first-order error-propagation argument, that R2 cancels the parallel component of mean-estimation error that R1 retains. Across 38 models on the Massive Multilingual Text Embedding Benchmark (MMTEB)~\\citep{MMTEB}, R2 yields consistent classification"},"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":"2511.11041","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2025-11-14T07:51:59Z","cross_cats_sorted":["cs.AI","cs.LG"],"title_canon_sha256":"19c743577b23c138682ab01906f609f70d868ed15149c1d8b4f0cd02afecd24c","abstract_canon_sha256":"8df17d2865d226056852e0105ebd292550d2305c3377f0dddc0e71b7a403532c"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-06-09T01:04:39.990625Z","signature_b64":"VltkeVSeBDjgjsTUjNPlrOKDpbe4K23YLBplOf+vi2pQCRdXFiWbuoKjFX5E4XTxb3LD3DJld+wqcfxwoh5MDA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"2561cb8454d6d7529b881f445071f3cdb92f14bf8dfe3ff5fc8f2ab89a04c899","last_reissued_at":"2026-06-09T01:04:39.990146Z","signature_status":"signed_v1","first_computed_at":"2026-06-09T01:04:39.990146Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Correcting Mean Bias in Text Embeddings: A Refined Renormalization with Training-Free Improvements on MMTEB","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","cs.LG"],"primary_cat":"cs.CL","authors_text":"Haoyu Liang, Xingyu Ren, Youran Sun","submitted_at":"2025-11-14T07:51:59Z","abstract_excerpt":"We find that current sentence-embedding models produce outputs with a consistent bias: every embedding $e$ decomposes as $\\tilde e + \\mu$, where the mean $\\mu$ is near-identical across all sentences. We study two training-free corrections -- subtracting $\\mu$ directly (R1), or projecting each embedding off the mean direction (R2) -- and show, via a first-order error-propagation argument, that R2 cancels the parallel component of mean-estimation error that R1 retains. Across 38 models on the Massive Multilingual Text Embedding Benchmark (MMTEB)~\\citep{MMTEB}, R2 yields consistent classification"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2511.11041","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/2511.11041/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":"2511.11041","created_at":"2026-06-09T01:04:39.990205+00:00"},{"alias_kind":"arxiv_version","alias_value":"2511.11041v2","created_at":"2026-06-09T01:04:39.990205+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2511.11041","created_at":"2026-06-09T01:04:39.990205+00:00"},{"alias_kind":"pith_short_12","alias_value":"EVQ4XBCU23LV","created_at":"2026-06-09T01:04:39.990205+00:00"},{"alias_kind":"pith_short_16","alias_value":"EVQ4XBCU23LVFG4I","created_at":"2026-06-09T01:04:39.990205+00:00"},{"alias_kind":"pith_short_8","alias_value":"EVQ4XBCU","created_at":"2026-06-09T01:04:39.990205+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/EVQ4XBCU23LVFG4ID5CFA4PTZW","json":"https://pith.science/pith/EVQ4XBCU23LVFG4ID5CFA4PTZW.json","graph_json":"https://pith.science/api/pith-number/EVQ4XBCU23LVFG4ID5CFA4PTZW/graph.json","events_json":"https://pith.science/api/pith-number/EVQ4XBCU23LVFG4ID5CFA4PTZW/events.json","paper":"https://pith.science/paper/EVQ4XBCU"},"agent_actions":{"view_html":"https://pith.science/pith/EVQ4XBCU23LVFG4ID5CFA4PTZW","download_json":"https://pith.science/pith/EVQ4XBCU23LVFG4ID5CFA4PTZW.json","view_paper":"https://pith.science/paper/EVQ4XBCU","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2511.11041&json=true","fetch_graph":"https://pith.science/api/pith-number/EVQ4XBCU23LVFG4ID5CFA4PTZW/graph.json","fetch_events":"https://pith.science/api/pith-number/EVQ4XBCU23LVFG4ID5CFA4PTZW/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/EVQ4XBCU23LVFG4ID5CFA4PTZW/action/timestamp_anchor","attest_storage":"https://pith.science/pith/EVQ4XBCU23LVFG4ID5CFA4PTZW/action/storage_attestation","attest_author":"https://pith.science/pith/EVQ4XBCU23LVFG4ID5CFA4PTZW/action/author_attestation","sign_citation":"https://pith.science/pith/EVQ4XBCU23LVFG4ID5CFA4PTZW/action/citation_signature","submit_replication":"https://pith.science/pith/EVQ4XBCU23LVFG4ID5CFA4PTZW/action/replication_record"}},"created_at":"2026-06-09T01:04:39.990205+00:00","updated_at":"2026-06-09T01:04:39.990205+00:00"}