generative-retrieval

[Pytorch] Generative retrieval model using semantic IDs from "Recommender Systems with Generative Retrieval"

[Pytorch] Unofficial Implementation of "Recommender Systems with Generative Retrieval"

Updating curated list of research advancements on item identification and item tokenization in generative recommender systems. The survey is titled "A Survey of Item Identifiers in Generative Recommendation: Construction, Alignment, and Generation"

Differentiable Semantic ID for Generative Recommendation

Official Implementation of GENIUS: A Generative Framework for Universal Multimodal Search, CVPR 2025

We introduce the direct document relevance optimization (DDRO) for training a pairwise ranker model. DDRO encourages the model to focus on document-level relevance during generation

Code for "Multi-level Relevance Document Identifier Learning for Generative Retrieval". ACL 2025.

Code for "HierGR: Hierarchical Semantic Representation Enhancement for Generative Retrieval in Food Delivery Search". ACL 2025.

Reproducibility study of Planning-Ahead Generative Retrieval (PAG), probing lexical planner robustness under query perturbations and cross-lingual evaluation. SIGIR 2026 Reproducibility Track.

[Pytorch] Efficient tokenization for recommendations and generative retrieval. Inspired by STATIC decoding from "Vectorizing the Trie"

An open-source implementation of ToC-Aware RAG inspired by IBM's STAIR paper. Preserves document hierarchy for structured addressable retrieval and near-zero hallucinations.

Generative-retrieval memory router (DSI: query→note-id) vs. embedding retrieval on a small, linked, franglais personal-memory corpus — with a hardened, confound-free eval. Reproducible.

Open reimplementation of STAIR: structure-aware generative retrieval that returns real section titles instead of chunk IDs. Constrained decoding gives 0.00% hallucination. Runs Mistral-7B QLoRA in 5GB.

The Integration of Chain-of-Thought in Generative Information Retrieval

Official code for ZeroGR: A Generalizable and Scalable Framework for Zero-Shot Generative Retrieval (arXiv:2510.10419)

Official code for "Generative Retrieval for Cover Song Identification via Audio-Derived Semantic IDs" (accepted at ICECCME 2026, IEEE)

A verified SASRec reproduction and a semantic-ID generative recommender on one backbone, one eval harness, one set of frozen negatives. Full-catalog vs. sampled ranking, a measured seed-noise floor, and every README table CI-asserted against the run that produced it.

A reproduction of OneRetrieval KAE.

Project for the University of Bologna "Cognitive Semantics for Artificial Intelligence" course (a.y. 2024-25).

SASRec с нуля на PyTorch (causal self-attention вручную) на Yandex Yambda. Аудио-эмбеддинги квантованы в semantic IDs через RQ-VAE: как замена item-эмбеддингов проигрывают, как дополнение дают +39% к NDCG@10 (0.575 vs 0.413). Ablation из шести моделей