VOOZH about

URL: https://arxiv.org/abs/2511.18659

⇱ [2511.18659] CLaRa: Bridging Retrieval and Generation with Continuous Latent Reasoning

Computer Science > Computation and Language

arXiv:2511.18659 (cs)
[Submitted on 24 Nov 2025 (v1), last revised 5 Feb 2026 (this version, v3)]

Title:CLaRa: Bridging Retrieval and Generation with Continuous Latent Reasoning

View PDF HTML (experimental)
Abstract: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.
Subjects: Computation and Language (cs.CL)
Cite as: arXiv:2511.18659 [cs.CL]
(or arXiv:2511.18659v3 [cs.CL] for this version)
https://doi.org/10.48550/arXiv.2511.18659
arXiv-issued DOI via DataCite

Submission history

From: Jie He [view email]
[v1] Mon, 24 Nov 2025 00:11:14 UTC (20,087 KB)
[v2] Tue, 25 Nov 2025 22:02:20 UTC (22,521 KB)
[v3] Thu, 5 Feb 2026 19:03:09 UTC (22,489 KB)
Full-text links:

Access Paper:

👁 license icon
view license
Current browse context:
cs.CL
< prev   |   next >
Change to browse by:
cs

References & Citations

export BibTeX citation

BibTeX formatted citation

×
Data provided by:

Bibliographic and Citation Tools

Bibliographic Explorer (What is the Explorer?)
Connected Papers (What is Connected Papers?)
Litmaps (What is Litmaps?)
scite Smart Citations (What are Smart Citations?)

Code, Data and Media Associated with this Article

alphaXiv (What is alphaXiv?)
CatalyzeX Code Finder for Papers (What is CatalyzeX?)
DagsHub (What is DagsHub?)
Gotit.pub (What is GotitPub?)
Hugging Face (What is Huggingface?)
Papers with Code (What is Papers with Code?)
ScienceCast (What is ScienceCast?)

Demos

Replicate (What is Replicate?)
Hugging Face Spaces (What is Spaces?)
TXYZ.AI (What is TXYZ.AI?)

Recommenders and Search Tools

Influence Flower (What are Influence Flowers?)
CORE Recommender (What is CORE?)

arXivLabs: experimental projects with community collaborators

arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.

Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.

Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.