Translational Generative Retrieval via Potential Query Generation (ICASSP 2025)

This post summarizes an ICASSP 2025 paper: Translational Generative Retrieval via Potential Query Generation.

• Source: IEEE Xplore
• Type: Academic paper
• Area: Information retrieval / search and recommendation systems

Key points

• The paper proposes a translational generative retrieval approach based on potential query generation, targeting core IR challenges: efficiency, scalability, and user intent understanding in large-scale search, recommendation, and personalization systems.
• It responds to a central LLM-era question: how to redistribute responsibilities among retrieval, ranking, generation, and tool calling, moving beyond traditional pipelines that treat these stages in isolation.
• Relevant scenarios include open-domain information access, enterprise knowledge retrieval, conversational search, and end-to-end architectures that couple external knowledge sources with generative models.

Methodological framing

Typical systems in this space follow a four-step pipeline:

1. Input & representation — encode queries, documents, and user context as dense/sparse representations or structured prompts. 2. Core modules — retriever, reranker, planner, memory, and tool interfaces, chained or combined per task. 3. Learning strategy — supervised fine-tuning, contrastive learning, distillation, reinforcement learning (including process rewards), and synthetic data bootstrapping. 4. Inference strategy — single-pass retrieval, iterative retrieval, parallel sub-queries, early stopping, and budget control.

Evaluation context

Common evaluation setups for related work:

• Datasets: MS MARCO, BEIR, Natural Questions, domain corpora, recommendation benchmarks.
• Metrics: nDCG@10, MRR, Recall@k, Hit@k, human preference, task success rate, latency, and token cost.
• Baselines: BM25, dense retrieval, cross-encoder reranking, retrieval-free LLMs, commercial search APIs.

Specific quantitative results in this paper should be verified against the original PDF; this summary is based on the abstract and public metadata.

Insights for search / rec / personalization

1. Architecture: cascade retrieval + rerank + generation remains mainstream, but agentic paradigms are making *retrieval policy itself* a learnable object. 2. Data: high-quality instruction data and click/session logs matter equally; synthetic data requires safeguards against knowledge leakage and distribution shift. 3. Evaluation: the gap between offline metrics and online satisfaction is widening; LLM-as-judge needs cross-validation with human evaluation. 4. Production: latency, cost, interpretability, and safety are hard constraints — do not optimize only for academic benchmarks.

Engineering checklist

| Item | Question | Suggestion | |------|----------|------------| | Data | PII in training/index? Versioning? | Partitioned indexes, anonymization, rollback-capable embedding versions | | Latency | p99 budget? Retrieval steps? | Cascade + early stop, cache hot queries, async reranking | | Quality | Do offline gains translate online? | Interleaving experiments, human audits, citation checks | | Safety | Poisoning/bias from open retrieval? | Source whitelisting, adversarial detection, output filtering | | Cost | Per-query tokens / GPU usage? | Route to smaller models, distillation, hybrid sparse+dense |

Related entries

• A Comprehensive Study of Knowledge Editing for Large Language Models (arXiv:2401.01286)
• INTERS: Unlocking the Power of Large Language Models in Search (arXiv:2401.06532)
• Recommendation as Instruction Following (ACM)
• RouteLLM: Learning to Route LLMs with Preference Data (arXiv:2406.18665)
• Representation Learning with Large Language Models for Recommendation (ACM)

Glossary

| Term | Meaning | |------|---------| | IR | Information Retrieval | | RAG | Retrieval-Augmented Generation | | nDCG | Normalized Discounted Cumulative Gain | | Agentic Search | Search modeled as sequential decision-making and tool calling | | Gen-IR | Generative Information Retrieval |