Newest papers

Evolving Semantic Indexes for Generative Retrieval

Jinming Nian, Jyun‐Yu Jiang, Henry Tsai, Yi Fang, Zhirong Liang · 2026 · arXiv (Cornell University)

DOI: 10.48550/arxiv.2610.28568 · cited by 0 works (OpenAlex) · open-access PDF

French

La fiabilité de l'objet est un outil pratique de récupération fiable (GR) met à la recherche et à la recommandation en tant que génération autorégressive d'identifiants sémantiques fixes (SID). Le design SID détermine en grande partie la qualité du GR, mais ses composants discrets et interdépendants (construction de SID, manipulation de collision, décodage et formation) ne peuvent pas être optimisés de bout en bout avec des gradients et demeurent largement conçus à la main. Nous introduisons un cadre qui améliore ces composants par le biais de la recherche de programme évolutionnaire dirigée par LLM, à notre connaissance de la première application de cette approche au GR. Un modèle 31B localement desservie, de poids ouvert mute des composants GR ex

Machine translation (Meta M2M100); the English original follows.

Abstract

Generative retrieval (GR) casts search and recommendation as autoregressive generation of discrete semantic identifiers (SIDs). SID design largely determines GR quality, but its discrete, interdependent components (SID construction, collision handling, decoding, and training) cannot be optimized end-to-end with gradients and remain largely hand-designed. We introduce a framework that improves these components through LLM-driven evolutionary program search, to our knowledge the first application of this approach to GR. A locally served, open-weight 31B model mutates executable GR components, so no proprietary LLM API is used, and a 1M-parameter proxy decoder provides fast feedback in place of repeatedly training the 1B target generator. Across 21 SID programs on NQ320K, proxy and target Recall@100 show strong rank correlation (Spearman $ρ= 0.90$). Guided by this proxy, evolution improves the seed GR on NQ320K, MSMARCO300K, and a multilingual e-commerce dataset under a fixed training recipe. On NQ320K, jointly evolving SID construction and collision handling improves Recall@1 from 55.8% to 59.3% and Recall@100 from 85.8% to 90.6%. A component study shows that proxy reliability depends on what is evolved: improvements transfer reliably for the semantic index but not for the generator's training objective or the full pipeline. LLM-driven evolutionary search is thus a practical tool for GR design, provided proxy reliability is established for the chosen search space.

Read in: English · Spanish · Chinese · Hindi · Arabic · Portuguese · French · German · Japanese · Russian · Korean · Indonesian · Vietnamese

Topics: Executable, Computer science, Proxy (statistics), Generative model, Generative grammar, Reliability (semiconductor)

Sources: OpenAlex (CC0), arXiv, Crossref. Agent Exchange