{"paper":{"title":"QuIVer: Rethinking ANN Graph Topology via Training-Free Binary Quantization","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"QuIVer builds ANN graph indices entirely inside a 2-bit sign-magnitude binary quantization space.","cross_cats":[],"primary_cat":"cs.DB","authors_text":"Chengcheng Li, Wenxuan Xiao, Zhiyou Wang","submitted_at":"2026-05-04T03:04:12Z","abstract_excerpt":"Approximate nearest neighbor (ANN) graph indices such as HNSW and Vamana construct their edge topology in full-precision or high-fidelity quantized metric spaces, relegating binary quantization (BQ) to a post-hoc distance estimator during search. This paper asks a different question: Can binary quantization define the graph topology itself -- and if so, under what conditions? We study this question through QuIVer (Quantized Index for Vector Retrieval), a training-free ANN graph index that performs Vamana edge selection, diversity pruning, and beam-search navigation entirely within a 2-bit Sign"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"QuIVer performs edge selection, pruning, and graph navigation entirely in a 2-bit Sign-Magnitude BQ metric space, achieving at least 88% Recall@10 at 13--41K multi-threaded QPS with 58--262-second construction and less than 1.3 GB hot memory on six embedding datasets.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"That 2-bit sign-magnitude quantization preserves enough distance information for Vamana-style alpha-diversity pruning and beam search to produce a graph topology whose quality is close to full-precision construction, at least for cosine-native low-effective-dimensionality data.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"QuIVer constructs ANN graphs using only 2-bit sign-magnitude binary quantization for topology decisions, achieving at least 88% Recall@10 at high throughput with low memory on embedding datasets.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"QuIVer builds ANN graph indices entirely inside a 2-bit sign-magnitude binary quantization space.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"98227fa1ac77948c666d7009359c15384e32215c06edde6dc25d40a6a2384184"},"source":{"id":"2605.02171","kind":"arxiv","version":3},"verdict":{"id":"5a412625-a302-4e7e-97bf-6b9505cc9cac","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-12T01:48:47.909907Z","strongest_claim":"QuIVer performs edge selection, pruning, and graph navigation entirely in a 2-bit Sign-Magnitude BQ metric space, achieving at least 88% Recall@10 at 13--41K multi-threaded QPS with 58--262-second construction and less than 1.3 GB hot memory on six embedding datasets.","one_line_summary":"QuIVer constructs ANN graphs using only 2-bit sign-magnitude binary quantization for topology decisions, achieving at least 88% Recall@10 at high throughput with low memory on embedding datasets.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"That 2-bit sign-magnitude quantization preserves enough distance information for Vamana-style alpha-diversity pruning and beam search to produce a graph topology whose quality is close to full-precision construction, at least for cosine-native low-effective-dimensionality data.","pith_extraction_headline":"QuIVer builds ANN graph indices entirely inside a 2-bit sign-magnitude binary quantization space."},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2605.02171/integrity.json","findings":[],"available":true,"detectors_run":[{"name":"doi_compliance","ran_at":"2026-05-19T16:37:36.711643Z","status":"completed","version":"1.0.0","findings_count":0}],"snapshot_sha256":"c388a9c9ce25618f32a05a45abf3e810edb440d1007d64dc20d566b98e53cae0"},"references":{"count":0,"sample":[],"resolved_work":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57","internal_anchors":0},"formal_canon":{"evidence_count":2,"snapshot_sha256":"3361213e65725c136f33317cd532b624d9f62839080d35efa45253131963ecd1"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"}