# One Model Is Enough: Native Retrieval Embeddings from LLM Agent Hidden States

arXiv: https://arxiv.org/abs/2603.08429 · HTML: https://mlsanity.com/paper/2603.08429/

## Key facts

- Cited by 2 papers in the map (data through 14 Oct 2026).
- Placed in Agent memory systems, one of 16 directions in LLM agents and tool use.
- Builds on 9 of its 13 references that are in the map.
- First posted on arXiv 9 Mar 2026.

## Placed in

- LLM agents and tool use / [Agent memory systems](https://mlsanity.com/agents/agent-memory-systems/)

## Builds on (most cited first, 9 of 9)

- [Distilling the Knowledge in a Neural Network](https://mlsanity.com/paper/1503.02531/) (2015)
- [Retrieval-Augmented Generation for Knowledge-Intensive NLP Tasks](https://mlsanity.com/paper/2005.11401/) (2020)
- [Sentence-BERT: Sentence Embeddings using Siamese BERT-Networks](https://mlsanity.com/paper/1908.10084/) (2019)
- [Representation Learning with Contrastive Predictive Coding](https://mlsanity.com/paper/1807.03748/) (2018)
- [Dense Passage Retrieval for Open-Domain Question Answering](https://mlsanity.com/paper/2004.04906/) (2020)
- [Precise Zero-Shot Dense Retrieval without Relevance Labels](https://mlsanity.com/paper/2212.10496/) (2022)
- [NV-Embed: Improved Techniques for Training LLMs as Generalist Embedding Models](https://mlsanity.com/paper/2405.17428/) (2024)
- [Open-Domain Question Answering Goes Conversational via Question Rewriting](https://mlsanity.com/paper/2010.04898/) (2020)
- [Improving Efficient Neural Ranking Models with Cross-Architecture Knowledge Distillation](https://mlsanity.com/paper/2010.02666/) (2020)

## Built on by (most cited first, 2 of 2)

- [Lightweight Chunk Selection for Mobile Retrieval-Augmented Generation](https://mlsanity.com/paper/2608.03148/) (2026)
- [RALI@iKAT 2026: Retrieval-aware User Profile Denoising via Personalized Conversational Dense Retrieval](https://mlsanity.com/paper/2610.17673/) (2026)

## Abstract (arXiv)

LLM agents that retrieve external knowledge typically generate a search query as text, then run a separate embedding model to encode it into a vector. This two-model pipeline adds infrastructure complexity and latency, yet is redundant: the LLM already encodes the full conversational context in its hidden states. We propose equipping LLM agents with native retrieval capability by adding a lightweight projection head that maps hidden states directly into the embedding space, eliminating the need for a separate embedding model. Trained with a combination of alignment, contrastive, and rank distillation losses, our method retains 97% of baseline retrieval quality while enabling the LLM agent to search with its own representations. Experiments on the QReCC conversational search benchmark show competitive Recall@10 and MRR@10 compared to the standard generate-then-encode pipeline, with systematic ablations confirming the contribution of each loss component.