{"paper":{"title":"Query-efficient model evaluation using cached responses","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"DKPS with cached responses allows benchmark evaluation of new models using far fewer queries while matching baseline accuracy.","cross_cats":["cs.AI","stat.ME"],"primary_cat":"cs.LG","authors_text":"Ben Johnson, Carey Priebe, Hayden Helm","submitted_at":"2026-05-08T01:24:06Z","abstract_excerpt":"Evaluating a new model on an existing benchmark is often necessary to understand its behavior before deployment. For modern evaluation frameworks, generating and evaluating a response for all queries can be prohibitively expensive. In practice, responses from previously-evaluated models are often cached -- creating a potential opportunity to use this additional information to decrease the number of queries required to accurately evaluate a new model. In this paper, we introduce an approach for predicting benchmark performance that leverages cached model responses based on the Data Kernel Persp"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"DKPS-based methods achieve the same mean absolute error as baselines with a substantially decreased query budget.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"That the Data Kernel Perspective Space reliably quantifies black-box relationships between models and that the stated theoretical conditions for query-efficiency translate to practical benchmark settings.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"DKPS-based methods leverage cached model responses to achieve equivalent benchmark prediction accuracy with substantially fewer queries than standard evaluation.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"DKPS with cached responses allows benchmark evaluation of new models using far fewer queries while matching baseline accuracy.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"4c072ff360ac02f2a8ee2b02b8c9cc2a4f60f0364dc67aab3ea0bc97a666147d"},"source":{"id":"2605.07096","kind":"arxiv","version":2},"verdict":{"id":"4e016987-7957-4e68-929c-91e0c52e784c","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-11T00:50:12.566768Z","strongest_claim":"DKPS-based methods achieve the same mean absolute error as baselines with a substantially decreased query budget.","one_line_summary":"DKPS-based methods leverage cached model responses to achieve equivalent benchmark prediction accuracy with substantially fewer queries than standard evaluation.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"That the Data Kernel Perspective Space reliably quantifies black-box relationships between models and that the stated theoretical conditions for query-efficiency translate to practical benchmark settings.","pith_extraction_headline":"DKPS with cached responses allows benchmark evaluation of new models using far fewer queries while matching baseline accuracy."},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2605.07096/integrity.json","findings":[],"available":true,"detectors_run":[{"name":"claim_evidence","ran_at":"2026-05-20T11:22:03.427819Z","status":"completed","version":"1.0.0","findings_count":0},{"name":"ai_meta_artifact","ran_at":"2026-05-20T06:36:30.857371Z","status":"completed","version":"1.0.0","findings_count":0},{"name":"doi_title_agreement","ran_at":"2026-05-19T17:31:18.589559Z","status":"completed","version":"1.0.0","findings_count":0},{"name":"doi_compliance","ran_at":"2026-05-19T12:03:25.328571Z","status":"completed","version":"1.0.0","findings_count":0}],"snapshot_sha256":"cda9af02d44944d27322c7b56280a87d616ea44d08eea623d517094b72a544bd"},"references":{"count":0,"sample":[],"resolved_work":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57","internal_anchors":0},"formal_canon":{"evidence_count":2,"snapshot_sha256":"e372b0d3207c98a1cf5f2f12df769165ebf183137bd6defdff129a61f890384e"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"}