Search papers, labs, and topics across Lattice.
This paper introduces GRIP, a novel retrieval-augmented generation framework where retrieval decisions are integrated directly into the token-level decoding process of a language model. GRIP uses "Self-Triggered Information Planning" via control tokens to allow the model to dynamically decide when to retrieve, how to reformulate queries, and when to stop retrieving, all within a single autoregressive trajectory. Experiments on QA benchmarks demonstrate that GRIP outperforms strong RAG baselines and achieves performance competitive with GPT-4o while using significantly fewer parameters.
Forget external retrieval controllers: GRIP lets your language model decide when and how to retrieve information, all within its own token-level decoding process.
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 \textbf{GRIP} (\textbf{G}eneration-guided \textbf{R}etrieval with \textbf{I}nformation \textbf{P}lanning), a unified framework in which the model regulates retrieval behavior through control-token emission. Central to GRIP is \textit{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.
