{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:NQLWE7OF3D6VS4LVQTS7GQQBBB","short_pith_number":"pith:NQLWE7OF","schema_version":"1.0","canonical_sha256":"6c17627dc5d8fd59717584e5f3420108631fc265ea3aac745685e45efcadb664","source":{"kind":"arxiv","id":"2502.20379","version":1},"attestation_state":"computed","paper":{"title":"Multi-Agent Verification: Scaling Test-Time Compute with Multiple Verifiers","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.AI","authors_text":"Shalev Lifshitz, Sheila A. McIlraith, Yilun Du","submitted_at":"2025-02-27T18:53:30Z","abstract_excerpt":"By utilizing more computational resources at test-time, large language models (LLMs) can improve without additional training. One common strategy uses verifiers to evaluate candidate outputs. In this work, we propose a novel scaling dimension for test-time compute: scaling the number of verifiers. We introduce Multi-Agent Verification (MAV) as a test-time compute paradigm that combines multiple verifiers to improve performance. We propose using Aspect Verifiers (AVs), off-the-shelf LLMs prompted to verify different aspects of outputs, as one possible choice for the verifiers in a MAV system. A"},"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":"2502.20379","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.AI","submitted_at":"2025-02-27T18:53:30Z","cross_cats_sorted":[],"title_canon_sha256":"d4259c872614fb420c6d79864f552805cc4ad6702098fbdc8e2ece5e203f3a54","abstract_canon_sha256":"d595c6874caab7aa918affc0ef5e38ea95d32a9ac5fb72157169bbac9a945788"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T10:21:13.996591Z","signature_b64":"oknfa0jKe//fNd/9EftafQXFHVlt43DBvM2+xDmeNBUd2M3eJpoooCYxwqxbMjL6T4J6cNvtk0QEbWmpyuFHBw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"6c17627dc5d8fd59717584e5f3420108631fc265ea3aac745685e45efcadb664","last_reissued_at":"2026-07-05T10:21:13.996052Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T10:21:13.996052Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Multi-Agent Verification: Scaling Test-Time Compute with Multiple Verifiers","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.AI","authors_text":"Shalev Lifshitz, Sheila A. McIlraith, Yilun Du","submitted_at":"2025-02-27T18:53:30Z","abstract_excerpt":"By utilizing more computational resources at test-time, large language models (LLMs) can improve without additional training. One common strategy uses verifiers to evaluate candidate outputs. In this work, we propose a novel scaling dimension for test-time compute: scaling the number of verifiers. We introduce Multi-Agent Verification (MAV) as a test-time compute paradigm that combines multiple verifiers to improve performance. We propose using Aspect Verifiers (AVs), off-the-shelf LLMs prompted to verify different aspects of outputs, as one possible choice for the verifiers in a MAV system. A"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2502.20379","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/2502.20379/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":"2502.20379","created_at":"2026-07-05T10:21:13.996118+00:00"},{"alias_kind":"arxiv_version","alias_value":"2502.20379v1","created_at":"2026-07-05T10:21:13.996118+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2502.20379","created_at":"2026-07-05T10:21:13.996118+00:00"},{"alias_kind":"pith_short_12","alias_value":"NQLWE7OF3D6V","created_at":"2026-07-05T10:21:13.996118+00:00"},{"alias_kind":"pith_short_16","alias_value":"NQLWE7OF3D6VS4LV","created_at":"2026-07-05T10:21:13.996118+00:00"},{"alias_kind":"pith_short_8","alias_value":"NQLWE7OF","created_at":"2026-07-05T10:21:13.996118+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":10,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.07834","citing_title":"Cherry-pick Override: Unsafe Directional Commitment in LLM Judges under Mixed Evidence","ref_index":6,"is_internal_anchor":false},{"citing_arxiv_id":"2606.02859","citing_title":"Economy of Minds: Emerging Multi-Agent Intelligence with Economic Interactions","ref_index":23,"is_internal_anchor":false},{"citing_arxiv_id":"2606.01075","citing_title":"On the Generalization Gap in Self-Evolving Language Model Reasoning","ref_index":23,"is_internal_anchor":false},{"citing_arxiv_id":"2605.22138","citing_title":"Efficient Agentic Reasoning Through Self-Regulated Simulative Planning","ref_index":120,"is_internal_anchor":false},{"citing_arxiv_id":"2605.22154","citing_title":"IdleSpec: Exploiting Idle Time via Speculative Planning for LLM Agents","ref_index":22,"is_internal_anchor":false},{"citing_arxiv_id":"2601.21972","citing_title":"Learning Decentralized LLM Collaboration with Multi-Agent Actor Critic","ref_index":20,"is_internal_anchor":false},{"citing_arxiv_id":"2604.18547","citing_title":"FUSE: Ensembling Verifiers with Zero Labeled Data","ref_index":10,"is_internal_anchor":false},{"citing_arxiv_id":"2604.16029","citing_title":"Cut Your Losses! Learning to Prune Paths Early for Efficient Parallel Reasoning","ref_index":15,"is_internal_anchor":false},{"citing_arxiv_id":"2604.17112","citing_title":"Complementing Self-Consistency with Cross-Model Disagreement for Uncertainty Quantification","ref_index":23,"is_internal_anchor":false},{"citing_arxiv_id":"2604.19049","citing_title":"Refute-or-Promote: An Adversarial Stage-Gated Multi-Agent Review Methodology for High-Precision LLM-Assisted Defect Discovery","ref_index":14,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/NQLWE7OF3D6VS4LVQTS7GQQBBB","json":"https://pith.science/pith/NQLWE7OF3D6VS4LVQTS7GQQBBB.json","graph_json":"https://pith.science/api/pith-number/NQLWE7OF3D6VS4LVQTS7GQQBBB/graph.json","events_json":"https://pith.science/api/pith-number/NQLWE7OF3D6VS4LVQTS7GQQBBB/events.json","paper":"https://pith.science/paper/NQLWE7OF"},"agent_actions":{"view_html":"https://pith.science/pith/NQLWE7OF3D6VS4LVQTS7GQQBBB","download_json":"https://pith.science/pith/NQLWE7OF3D6VS4LVQTS7GQQBBB.json","view_paper":"https://pith.science/paper/NQLWE7OF","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2502.20379&json=true","fetch_graph":"https://pith.science/api/pith-number/NQLWE7OF3D6VS4LVQTS7GQQBBB/graph.json","fetch_events":"https://pith.science/api/pith-number/NQLWE7OF3D6VS4LVQTS7GQQBBB/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/NQLWE7OF3D6VS4LVQTS7GQQBBB/action/timestamp_anchor","attest_storage":"https://pith.science/pith/NQLWE7OF3D6VS4LVQTS7GQQBBB/action/storage_attestation","attest_author":"https://pith.science/pith/NQLWE7OF3D6VS4LVQTS7GQQBBB/action/author_attestation","sign_citation":"https://pith.science/pith/NQLWE7OF3D6VS4LVQTS7GQQBBB/action/citation_signature","submit_replication":"https://pith.science/pith/NQLWE7OF3D6VS4LVQTS7GQQBBB/action/replication_record"}},"created_at":"2026-07-05T10:21:13.996118+00:00","updated_at":"2026-07-05T10:21:13.996118+00:00"}