{"paper":{"title":"AI Research Preference Models","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.AI","authors_text":"Alberto Pepe, Alexander D. Goldie, Anirudh Goyal, Anya Sims, Bassel Al Omari, Bhavul Gauri, Carl Domond, Daniel Izcovich, Despoina Magka, Emily McMilin, Eryk Helenowski, Hela Momand, Jakob Nicolaus Foerster, Jason Weston, Jean-Christophe Gagnon-Audet, Jo\\~ao Henriques, Karen Hambardzumyan, Lucia Cipolina-Kun, Martin Josifoski, Masoud Jalili Sabet, Muna Aghamelu, Nicolas Baldwin, Noam Levi, Paris Giampouras, Rishi Hazra, Saba Nazir, Tatiana Shavrina, Thomas Mann, Thomas Simon Foster, Tingchen Fu, Xian Li, Yoram Bachrach, Yulin Wang","submitted_at":"2026-08-14T04:20:37Z","abstract_excerpt":"AI research agents (AIRA) can now propose, implement, and evaluate their own machine learning experiments, but progress on frontier tasks is throttled by cost: a candidate solution can be written in minutes, whereas evaluating it can take hours to days of GPU time. An agent can therefore propose far more candidates than it can afford to run, and its progress depends on its research preference: how it allocates a fixed execution budget across many candidates. We introduce AI Research Preference Models (RPMs) that predict which of multiple candidate solutions are most worth executing, without pa"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2608.13940","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/2608.13940/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"}