mlsanityData to 14 Oct 2026

CLaRa: Bridging Retrieval and Generation with Continuous Latent Reasoning

arXiv 2511.18659 ↗First posted 24 Nov 2025cs.CLcited by 11

On the map: LLM reasoning → Knowledge- and tool-augmented reasoning

also in 1 more
Abstract (arXiv)

Retrieval-augmented generation (RAG) enhances large language models (LLMs) with external knowledge but still suffers from long contexts and disjoint retrieval-generation optimization. In this work, we propose CLaRa (Continuous Latent Reasoning), a unified framework that performs embedding-based compression and joint optimization in a shared continuous space. To obtain semantically rich and retrievable compressed vectors, thereby reducing the document length fed into the generator, we introduce SCP, a key-preserving data synthesis framework based on question answering and paraphrase supervision. CLaRa then trains the reranker and generator end-to-end via a single language modeling loss, with gradients flowing through both modules using a differentiable top-k estimator. Theoretically, this unified optimization aligns retrieval relevance with answer quality. Experiments across multiple QA benchmarks show that CLaRa achieves state-of-the-art compression and reranking performance, even at a text compression rate of 16, outperforming text-based fine-tuned baselines.

Key facts. Cited by 11 papers in the map (data through 14 Oct 2026). Placed in Knowledge- and tool-augmented reasoning, one of 17 directions in LLM reasoning. Builds on 44 of its 53 references that are in the map. First posted on arXiv 24 Nov 2025.

Builds on

44 of 53 references.

Closest ancestors · 3

Same direction, newest first

  1. 2025Search-R1: Training LLMs to Reason and Leverage Search Engines with Reinforcement Learning12 Mar
  2. 2025Rank-R1: Enhancing Reasoning in LLM-based Document Rerankers via Reinforcement Learning8 Mar
  3. 2023Retrieval-Augmented Generation for Large Language Models: A Survey18 Dec

Other references · 41

  1. 2020Retrieval-Augmented Generation for Knowledge-Intensive NLP Taskscited by 9,297
  2. 2019BERTScore: Evaluating Text Generation with BERTcited by 5,512
  3. 2018HotpotQA: A Dataset for Diverse, Explainable Multi-hop Question Answeringcited by 3,754
  4. 2013Estimating or Propagating Gradients Through Stochastic Neurons for Conditional Computationcited by 2,079
  5. 2023Self-RAG: Learning to Retrieve, Generate, and Critique through Self-Reflectioncited by 1,839
  6. 2020Constructing A Multi-hop QA Dataset for Comprehensive Evaluation of Reasoning Stepscited by 1,282
  7. 2022Atlas: Few-shot Learning with Retrieval Augmented Language Modelscited by 952
  8. 2021MuSiQue: Multihop Questions via Single-hop Question Compositioncited by 916
  9. 2024Training Large Language Models to Reason in a Continuous Latent Spacecited by 797
  10. 2023Learning to Compress Prompts with Gist Tokenscited by 390
  11. 2023Making Retrieval-Augmented Language Models Robust to Irrelevant Contextcited by 369
  12. 2023Adapting Language Models to Compress Contextscited by 356
29 more
  1. 2024LLMLingua-2: Data Distillation for Efficient and Faithful Task-Agnostic Prompt Compressioncited by 348
  2. 2024Tool Learning with Large Language Models: A Surveycited by 223
  3. 2024RankRAG: Unifying Context Ranking with Retrieval-Augmented Generation in LLMscited by 202
  4. 2023In-context Autoencoder for Context Compression in a Large Language Modelcited by 199
  5. 2023RA-DIT: Retrieval-Augmented Dual Instruction Tuningcited by 195
  6. 2023RankZephyr: Effective and Robust Zero-Shot Listwise Reranking is a Breeze!cited by 155
  7. 2021End-to-End Training of Multi-Document Reader and Retriever for Open-Domain Question Answeringcited by 133
  8. 2024xRAG: Extreme Context Compression for Retrieval-augmented Generation with One Tokencited by 125
  9. 2024Retrieval Augmented Generation or Long-Context LLMs? A Comprehensive Study and Hybrid Approachcited by 119
  10. 2024Reducing hallucination in structured outputs via Retrieval-Augmented Generationcited by 93
  11. 2024CodeRAG-Bench: Can Retrieval Augment Code Generation?cited by 91
  12. 2024ChatQA: Surpassing GPT-4 on Conversational QA and RAGcited by 90
  13. 2024Retrieval-Augmented Generation for Natural Language Processing: A Surveycited by 71
  14. 2024Inference Scaling for Long-Context Retrieval Augmented Generationcited by 65
  15. 2023A Setwise Approach for Effective and Highly Efficient Zero-shot Ranking with Large Language Modelscited by 55
  16. 2024Generate-then-Ground in Retrieval-Augmented Generation for Multi-hop Question Answeringcited by 51
  17. 2024Don't Forget to Connect! Improving RAG with Graph-based Rerankingcited by 49
  18. 2025Ask in Any Modality: A Comprehensive Survey on Multimodal Retrieval-Augmented Generationcited by 42
  19. 2025MetaEmbed: Scaling Multimodal Retrieval at Test-Time with Flexible Late Interactioncited by 39
  20. 2024RAG-DDR: Optimizing Retrieval-Augmented Generation Using Differentiable Data Rewardscited by 33
  21. 2025Cramming 1568 Tokens into a Single Vector and Back Again: Exploring the Limits of Embedding Space Capacitycited by 28
  22. 2024Understand What LLM Needs: Dual Preference Alignment for Retrieval-Augmented Generationcited by 27
  23. 2024Long Context RAG Performance of Large Language Modelscited by 23
  24. 2025PISCO: Pretty Simple Compression for Retrieval-Augmented Generationcited by 20
  25. 2024Superposition Prompting: Improving and Accelerating Retrieval-Augmented Generationcited by 15
  26. 2025Understanding and Improving Information Preservation in Prompt Compression for LLMscited by 15
  27. 2025Direct Retrieval-augmented Optimization: Synergizing Knowledge Selection and Language Modelscited by 9
  28. 2025Reading Between the Timelines: RAG for Answering Diachronic Questionscited by 4
  29. 2025Gumbel Reranking: Differentiable End-to-End Reranker Optimizationcited by 2

Built on by

Cited by year (2026 to 14 Oct)

Citing papers per year: 2026: 11.

11 papers.

  1. 2026The Latent Space: Foundation, Evolution, Mechanism, Ability, and Outlookcited by 43
  2. 2026Scaling Video Understanding via Compact Latent Multi-Agent Collaborationcited by 4
  3. 2026LLM2Vec-Gen: Generative Embeddings from Large Language Modelscited by 2
  4. 2026Cognis: Context-Aware Memory for Conversational AI Agentscited by 1
  5. 2026A Vision for Multisensory Intelligence: Sensing, Science, and Synergycited by 0
  6. 2026LatentRAG: Latent Reasoning and Retrieval for Efficient Agentic RAGcited by 0
  7. 2026One Token per Multimodal Evidence: Latent Memory for Resource-Constrained QAcited by 0
  8. 2026SproutRAG: Attention-Guided Tree Search with Progressive Embeddings for Long-Document RAGcited by 0
  9. 2026MCompassRAG: Topic Metadata as a Semantic Compass for Paragraph-Level Retrievalcited by 0
  10. 2026Token-Native Storage: Read and Write in your Agent's Languagecited by 0
  11. 2026MaSRead: Content-Addressed Reading of Replicated Latent Storescited by 0

Similar, not cited

Close in wording, no citation either way

  1. 2026Latent Abstraction for Retrieval-Augmented Generation
  2. 2025SARA: Selective and Adaptive Retrieval-augmented Generation with Context Compression
  3. 2025ELITE: Embedding-Less retrieval with Iterative Text Exploration
  4. 2025Retrieval-Augmented Generation: A Comprehensive Survey of Architectures, Enhancements, and Robustness Frontiers
  5. 2024RE-RAG: Improving Open-Domain QA Performance and Interpretability with Relevance Estimator in Retrieval-Augmented Generation
  6. 2025LIR3AG: A Lightweight Rerank Reasoning Strategy Framework for Retrieval-Augmented Generation
  7. 2024Contextual Compression in Retrieval-Augmented Generation for Large Language Models: A Survey
  8. 2024BRIEF: Bridging Retrieval and Inference for Multi-hop Reasoning via Compression