{"bundle_type":"pith_open_graph_bundle","bundle_version":"1.0","pith_number":"pith:2026:5NVWHEFKHZ42IGKMLREXRZTWD5","short_pith_number":"pith:5NVWHEFK","canonical_record":{"source":{"id":"2604.22722","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.IR","submitted_at":"2026-04-24T17:18:56Z","cross_cats_sorted":["cs.AI","cs.LG"],"title_canon_sha256":"09a5078c4c8f9166a159b8242805126dd5e85a7a488361310384b5de4a7239a1","abstract_canon_sha256":"ac2be16751933f1624f4a66c96b0afe174a07b4885682c7e3fba00cc273599e9"},"schema_version":"1.0"},"canonical_sha256":"eb6b6390aa3e79a4194c5c4978e6761f4df3ed58fac3de6a0d27521a6327f2d9","source":{"kind":"arxiv","id":"2604.22722","version":2},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2604.22722","created_at":"2026-06-01T01:03:53Z"},{"alias_kind":"arxiv_version","alias_value":"2604.22722v2","created_at":"2026-06-01T01:03:53Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2604.22722","created_at":"2026-06-01T01:03:53Z"},{"alias_kind":"pith_short_12","alias_value":"5NVWHEFKHZ42","created_at":"2026-06-01T01:03:53Z"},{"alias_kind":"pith_short_16","alias_value":"5NVWHEFKHZ42IGKM","created_at":"2026-06-01T01:03:53Z"},{"alias_kind":"pith_short_8","alias_value":"5NVWHEFK","created_at":"2026-06-01T01:03:53Z"}],"events":[{"event_type":"record_created","subject_pith_number":"pith:2026:5NVWHEFKHZ42IGKMLREXRZTWD5","target":"record","payload":{"canonical_record":{"source":{"id":"2604.22722","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.IR","submitted_at":"2026-04-24T17:18:56Z","cross_cats_sorted":["cs.AI","cs.LG"],"title_canon_sha256":"09a5078c4c8f9166a159b8242805126dd5e85a7a488361310384b5de4a7239a1","abstract_canon_sha256":"ac2be16751933f1624f4a66c96b0afe174a07b4885682c7e3fba00cc273599e9"},"schema_version":"1.0"},"canonical_sha256":"eb6b6390aa3e79a4194c5c4978e6761f4df3ed58fac3de6a0d27521a6327f2d9","receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-06-01T01:03:53.619803Z","signature_b64":"Q902rVqPh/1F4I8SXhyUEh7t8TzKcdO4yX6GqECrXlZSCWpxtNNxr7EuoCdxoczMJ2IidcjAsqoKDoEAcyEyDA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"eb6b6390aa3e79a4194c5c4978e6761f4df3ed58fac3de6a0d27521a6327f2d9","last_reissued_at":"2026-06-01T01:03:53.618688Z","signature_status":"signed_v1","first_computed_at":"2026-06-01T01:03:53.618688Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"source_kind":"arxiv","source_id":"2604.22722","source_version":2,"attestation_state":"computed"},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-06-01T01:03:53Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"UglDzyHxjkcRqaEVyDa2wKTZGLIQ1XKKW+tSa0Mz5qEe5Sj1U+lNvJwtiDgzVmlKQaBOVNR4034HQrXF9StZAQ==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-04T22:29:37.208088Z"},"content_sha256":"b573ddfae91c0ae3d745b9605ef40cb3f731ef32b3281d9c7d4ed718c6fbf883","schema_version":"1.0","event_id":"sha256:b573ddfae91c0ae3d745b9605ef40cb3f731ef32b3281d9c7d4ed718c6fbf883"},{"event_type":"graph_snapshot","subject_pith_number":"pith:2026:5NVWHEFKHZ42IGKMLREXRZTWD5","target":"graph","payload":{"graph_snapshot":{"paper":{"title":"Aligning Dense Retrievers with LLM Utility via Distillation","license":"http://creativecommons.org/licenses/by/4.0/","headline":"Training a bi-encoder to match LLM perplexity utility distributions produces dense retrieval that is more accurate and 180 times faster than re-ranking.","cross_cats":["cs.AI","cs.LG"],"primary_cat":"cs.IR","authors_text":"Cheng Chang, Di Mu, Ga Wu, Himanshu Rai, Maksims Volkovs, Md Shahriar Tasjid, Rajinder Sandhu","submitted_at":"2026-04-24T17:18:56Z","abstract_excerpt":"Dense vector retrieval is the practical backbone of Retrieval- Augmented Generation (RAG), but similarity search can suffer from precision limitations. Conversely, utility-based approaches leveraging LLM re-ranking often achieve superior performance but are computationally prohibitive and prone to noise inherent in perplexity estimation. We propose Utility-Aligned Embeddings (UAE), a framework designed to merge these advantages into a practical, high-performance retrieval method. We formulate retrieval as a distribution matching problem, training a bi-encoder to imitate a utility distribution "},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"On the QASPER benchmark, UAE improves retrieval Recall@1 by 30.59%, MAP by 30.16% and Token F1 by 17.3% over the strong semantic baseline BGE-Base. UAE is over 180x faster than the efficient LLM re-ranking methods preserving competitive performance.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"The assumption that a bi-encoder can effectively imitate the utility distribution derived from LLM perplexity reduction without significant loss of information or introduction of biases from the distillation process.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"UAE trains bi-encoder retrievers to match LLM utility distributions via Utility-Modulated InfoNCE, yielding over 30% gains in Recall@1 and MAP on QASPER while running 180x faster than re-ranking.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"Training a bi-encoder to match LLM perplexity utility distributions produces dense retrieval that is more accurate and 180 times faster than re-ranking.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"ee412084a43c06ca66840b39bf9ca8d4463cef249bf88dc8733d18ed775cd97f"},"source":{"id":"2604.22722","kind":"arxiv","version":2},"verdict":{"id":"84bbc310-8df6-4cf3-b0e4-3c682e4e0900","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-08T10:10:12.463255Z","strongest_claim":"On the QASPER benchmark, UAE improves retrieval Recall@1 by 30.59%, MAP by 30.16% and Token F1 by 17.3% over the strong semantic baseline BGE-Base. UAE is over 180x faster than the efficient LLM re-ranking methods preserving competitive performance.","one_line_summary":"UAE trains bi-encoder retrievers to match LLM utility distributions via Utility-Modulated InfoNCE, yielding over 30% gains in Recall@1 and MAP on QASPER while running 180x faster than re-ranking.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"The assumption that a bi-encoder can effectively imitate the utility distribution derived from LLM perplexity reduction without significant loss of information or introduction of biases from the distillation process.","pith_extraction_headline":"Training a bi-encoder to match LLM perplexity utility distributions produces dense retrieval that is more accurate and 180 times faster than re-ranking."},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2604.22722/integrity.json","findings":[],"available":true,"detectors_run":[{"name":"ai_meta_artifact","ran_at":"2026-05-21T10:33:53.027343Z","status":"completed","version":"1.0.0","findings_count":0},{"name":"doi_compliance","ran_at":"2026-05-19T23:43:18.131855Z","status":"completed","version":"1.0.0","findings_count":0}],"snapshot_sha256":"4f883241244b2587456d6889a969e07805483996b5c6e66eebb05abc08466951"},"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"},"verdict_id":"84bbc310-8df6-4cf3-b0e4-3c682e4e0900"},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-06-01T01:03:53Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"+vQkijh5ptLYQkhxdsvmxxmaihVc3bybk1u4RdPrfdROl3iRmU1UEWl7xi3Jq0I8xXSYORR4ztJ6HvNGVOpvCA==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-04T22:29:37.208844Z"},"content_sha256":"9c4231ced7586f744ccbd179f25dbc8bca963479323a9b43e385d1968913aeaa","schema_version":"1.0","event_id":"sha256:9c4231ced7586f744ccbd179f25dbc8bca963479323a9b43e385d1968913aeaa"}],"timestamp_proofs":[],"mirror_hints":[{"mirror_type":"https","name":"Pith Resolver","base_url":"https://pith.science","bundle_url":"https://pith.science/pith/5NVWHEFKHZ42IGKMLREXRZTWD5/bundle.json","state_url":"https://pith.science/pith/5NVWHEFKHZ42IGKMLREXRZTWD5/state.json","well_known_bundle_url":"https://pith.science/.well-known/pith/5NVWHEFKHZ42IGKMLREXRZTWD5/bundle.json","status":"primary"}],"public_keys":[{"key_id":"pith-v1-2026-05","algorithm":"ed25519","format":"raw","public_key_b64":"stVStoiQhXFxp4s2pdzPNoqVNBMojDU/fJ2db5S3CbM=","public_key_hex":"b2d552b68890857171a78b36a5dccf368a953413288c353f7c9d9d6f94b709b3","fingerprint_sha256_b32_first128bits":"RVFV5Z2OI2J3ZUO7ERDEBCYNKS","fingerprint_sha256_hex":"8d4b5ee74e4693bcd1df2446408b0d54","rotates_at":null,"url":"https://pith.science/pith-signing-key.json","notes":"Pith uses this Ed25519 key to sign canonical record SHA-256 digests. Verify with: ed25519_verify(public_key, message=canonical_sha256_bytes, signature=base64decode(signature_b64))."}],"merge_version":"pith-open-graph-merge-v1","built_at":"2026-08-04T22:29:37Z","links":{"resolver":"https://pith.science/pith/5NVWHEFKHZ42IGKMLREXRZTWD5","bundle":"https://pith.science/pith/5NVWHEFKHZ42IGKMLREXRZTWD5/bundle.json","state":"https://pith.science/pith/5NVWHEFKHZ42IGKMLREXRZTWD5/state.json","well_known_bundle":"https://pith.science/.well-known/pith/5NVWHEFKHZ42IGKMLREXRZTWD5/bundle.json"},"state":{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2026:5NVWHEFKHZ42IGKMLREXRZTWD5","merge_version":"pith-open-graph-merge-v1","event_count":2,"valid_event_count":2,"invalid_event_count":0,"equivocation_count":0,"current":{"canonical_record":{"metadata":{"abstract_canon_sha256":"ac2be16751933f1624f4a66c96b0afe174a07b4885682c7e3fba00cc273599e9","cross_cats_sorted":["cs.AI","cs.LG"],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.IR","submitted_at":"2026-04-24T17:18:56Z","title_canon_sha256":"09a5078c4c8f9166a159b8242805126dd5e85a7a488361310384b5de4a7239a1"},"schema_version":"1.0","source":{"id":"2604.22722","kind":"arxiv","version":2}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2604.22722","created_at":"2026-06-01T01:03:53Z"},{"alias_kind":"arxiv_version","alias_value":"2604.22722v2","created_at":"2026-06-01T01:03:53Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2604.22722","created_at":"2026-06-01T01:03:53Z"},{"alias_kind":"pith_short_12","alias_value":"5NVWHEFKHZ42","created_at":"2026-06-01T01:03:53Z"},{"alias_kind":"pith_short_16","alias_value":"5NVWHEFKHZ42IGKM","created_at":"2026-06-01T01:03:53Z"},{"alias_kind":"pith_short_8","alias_value":"5NVWHEFK","created_at":"2026-06-01T01:03:53Z"}],"graph_snapshots":[{"event_id":"sha256:9c4231ced7586f744ccbd179f25dbc8bca963479323a9b43e385d1968913aeaa","target":"graph","created_at":"2026-06-01T01:03:53Z","signer":{"key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signer_id":"pith.science","signer_type":"pith_registry"},"payload":{"graph_snapshot":{"author_claims":{"count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57","strong_count":0},"builder_version":"pith-number-builder-2026-05-17-v1","claims":{"count":4,"items":[{"attestation":"unclaimed","claim_id":"C1","kind":"strongest_claim","source":"verdict.strongest_claim","status":"machine_extracted","text":"On the QASPER benchmark, UAE improves retrieval Recall@1 by 30.59%, MAP by 30.16% and Token F1 by 17.3% over the strong semantic baseline BGE-Base. UAE is over 180x faster than the efficient LLM re-ranking methods preserving competitive performance."},{"attestation":"unclaimed","claim_id":"C2","kind":"weakest_assumption","source":"verdict.weakest_assumption","status":"machine_extracted","text":"The assumption that a bi-encoder can effectively imitate the utility distribution derived from LLM perplexity reduction without significant loss of information or introduction of biases from the distillation process."},{"attestation":"unclaimed","claim_id":"C3","kind":"one_line_summary","source":"verdict.one_line_summary","status":"machine_extracted","text":"UAE trains bi-encoder retrievers to match LLM utility distributions via Utility-Modulated InfoNCE, yielding over 30% gains in Recall@1 and MAP on QASPER while running 180x faster than re-ranking."},{"attestation":"unclaimed","claim_id":"C4","kind":"headline","source":"verdict.pith_extraction.headline","status":"machine_extracted","text":"Training a bi-encoder to match LLM perplexity utility distributions produces dense retrieval that is more accurate and 180 times faster than re-ranking."}],"snapshot_sha256":"ee412084a43c06ca66840b39bf9ca8d4463cef249bf88dc8733d18ed775cd97f"},"formal_canon":{"evidence_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"integrity":{"available":true,"clean":true,"detectors_run":[{"findings_count":0,"name":"ai_meta_artifact","ran_at":"2026-05-21T10:33:53.027343Z","status":"completed","version":"1.0.0"},{"findings_count":0,"name":"doi_compliance","ran_at":"2026-05-19T23:43:18.131855Z","status":"completed","version":"1.0.0"}],"endpoint":"/pith/2604.22722/integrity.json","findings":[],"snapshot_sha256":"4f883241244b2587456d6889a969e07805483996b5c6e66eebb05abc08466951","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"Dense vector retrieval is the practical backbone of Retrieval- Augmented Generation (RAG), but similarity search can suffer from precision limitations. Conversely, utility-based approaches leveraging LLM re-ranking often achieve superior performance but are computationally prohibitive and prone to noise inherent in perplexity estimation. We propose Utility-Aligned Embeddings (UAE), a framework designed to merge these advantages into a practical, high-performance retrieval method. We formulate retrieval as a distribution matching problem, training a bi-encoder to imitate a utility distribution ","authors_text":"Cheng Chang, Di Mu, Ga Wu, Himanshu Rai, Maksims Volkovs, Md Shahriar Tasjid, Rajinder Sandhu","cross_cats":["cs.AI","cs.LG"],"headline":"Training a bi-encoder to match LLM perplexity utility distributions produces dense retrieval that is more accurate and 180 times faster than re-ranking.","license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.IR","submitted_at":"2026-04-24T17:18:56Z","title":"Aligning Dense Retrievers with LLM Utility via Distillation"},"references":{"count":0,"internal_anchors":0,"resolved_work":0,"sample":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2604.22722","kind":"arxiv","version":2},"verdict":{"created_at":"2026-05-08T10:10:12.463255Z","id":"84bbc310-8df6-4cf3-b0e4-3c682e4e0900","model_set":{"reader":"grok-4.3"},"one_line_summary":"UAE trains bi-encoder retrievers to match LLM utility distributions via Utility-Modulated InfoNCE, yielding over 30% gains in Recall@1 and MAP on QASPER while running 180x faster than re-ranking.","pipeline_version":"pith-pipeline@v0.9.0","pith_extraction_headline":"Training a bi-encoder to match LLM perplexity utility distributions produces dense retrieval that is more accurate and 180 times faster than re-ranking.","strongest_claim":"On the QASPER benchmark, UAE improves retrieval Recall@1 by 30.59%, MAP by 30.16% and Token F1 by 17.3% over the strong semantic baseline BGE-Base. UAE is over 180x faster than the efficient LLM re-ranking methods preserving competitive performance.","weakest_assumption":"The assumption that a bi-encoder can effectively imitate the utility distribution derived from LLM perplexity reduction without significant loss of information or introduction of biases from the distillation process."}},"verdict_id":"84bbc310-8df6-4cf3-b0e4-3c682e4e0900"}}],"author_attestations":[],"timestamp_anchors":[],"storage_attestations":[],"citation_signatures":[],"replication_records":[],"corrections":[],"mirror_hints":[],"record_created":{"event_id":"sha256:b573ddfae91c0ae3d745b9605ef40cb3f731ef32b3281d9c7d4ed718c6fbf883","target":"record","created_at":"2026-06-01T01:03:53Z","signer":{"key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signer_id":"pith.science","signer_type":"pith_registry"},"payload":{"attestation_state":"computed","canonical_record":{"metadata":{"abstract_canon_sha256":"ac2be16751933f1624f4a66c96b0afe174a07b4885682c7e3fba00cc273599e9","cross_cats_sorted":["cs.AI","cs.LG"],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.IR","submitted_at":"2026-04-24T17:18:56Z","title_canon_sha256":"09a5078c4c8f9166a159b8242805126dd5e85a7a488361310384b5de4a7239a1"},"schema_version":"1.0","source":{"id":"2604.22722","kind":"arxiv","version":2}},"canonical_sha256":"eb6b6390aa3e79a4194c5c4978e6761f4df3ed58fac3de6a0d27521a6327f2d9","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"eb6b6390aa3e79a4194c5c4978e6761f4df3ed58fac3de6a0d27521a6327f2d9","first_computed_at":"2026-06-01T01:03:53.618688Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-06-01T01:03:53.618688Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"Q902rVqPh/1F4I8SXhyUEh7t8TzKcdO4yX6GqECrXlZSCWpxtNNxr7EuoCdxoczMJ2IidcjAsqoKDoEAcyEyDA==","signature_status":"signed_v1","signed_at":"2026-06-01T01:03:53.619803Z","signed_message":"canonical_sha256_bytes"},"source_id":"2604.22722","source_kind":"arxiv","source_version":2}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:b573ddfae91c0ae3d745b9605ef40cb3f731ef32b3281d9c7d4ed718c6fbf883","sha256:9c4231ced7586f744ccbd179f25dbc8bca963479323a9b43e385d1968913aeaa"],"state_sha256":"e597f55f445a8478bf8d2bc80838f6c5829c09c58f36a9db2b1c56d699ce9477"},"bundle_signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"UNSUczPuDxxUBOTb4+N0BrHO8o5zKcHJX3AuUmXsxXdecRQoxdIJ3AU67s8mm1AxjNw+IHtajuDUdy+4m+rJAw==","signed_message":"bundle_sha256_bytes","signed_at":"2026-08-04T22:29:37.214537Z","bundle_sha256":"310976b195c2f8f54c92d5a0932f5d68bc87356d33b8208eb542b8c9c8aa468d"}}