Retrieval as Generation: A Unified Framework with Self-Triggered Information Planning

arXiv:2604.11407v1 Announce Type: cross Abstract: We revisit retrieval-augmented generation (RAG) by embedding retrieval control directly into generation. Instead of treating retrieval as an external intervention, we express retrieval decisions within token-level decoding, enabling end-to-end coordination without additional controllers or classifiers. Under the paradigm of Retrieval as Generation, we propose extbf{GRIP} (extbf{G}eneration-guided extbf{R}etrieval with extbf{I}nformation extbf{P}lanning), a unified framework in which the model regulates retrieval behavior through control-token emission. Central to GRIP is extit{Self-Triggered Information Planning}, which allows the model to decide when to retrieve, how to reformulate queries, and when to terminate, all within a single autoregressive trajectory. This design tightly couples retrieval and reasoning and supports dynamic multi-step inference with on-the-fly evidence integration. To supervise these behaviors, we construct a structured training set covering answerable, partially answerable, and multi-hop queries, each aligned with specific token patterns. Experiments on five QA benchmarks show that GRIP surpasses strong RAG baselines and is competitive with GPT-4o while using substantially fewer parameters.

Related

• Learning and Enforcing Context-Sensitive Control for LLMs
• Domain-Specific Data Generation Framework for RAG Adaptation
• CFMS: A Coarse-to-Fine Multimodal Synthesis Framework for Enhanced Tabular Reasoning
• Decomposing the Delta: What Do Models Actually Learn from Preference Pairs?
• SVSR: A Self-Verification and Self-Rectification Paradigm for Multimodal Reasoning

Source: arXiv cs.AI | 2026-04-14