Pith. sign in

REVIEW 18 cited by

Dynamic Parametric Retrieval Augmented Generation for Test-time Knowledge Enhancement

Not yet reviewed by Pith; the record is open.

This paper has not been read by Pith yet. Machine review is queued; the pith claim, tier, and objections will appear here once it completes.

SPECIMEN: schema-true, not a live event

T0 review · schema-true

One-sentence machine reading of the paper's core claim.

pith:XXXXXXXX · record.json · timestamp

arxiv 2503.23895 v4 pith:HKF27CVJ submitted 2025-03-31 cs.CL cs.AI

Dynamic Parametric Retrieval Augmented Generation for Test-time Knowledge Enhancement

classification cs.CL cs.AI
keywords knowledgeparametricllmscostsdypragtest-timeinferencetraining
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved
0 comments
Share X Bluesky LinkedIn Reddit HN
read the original abstract

Retrieval-augmented generation (RAG) enhances large language models (LLMs) by retrieving relevant documents from external sources and incorporating them into the context. While it improves reliability by providing factual texts, it significantly increases inference costs as context length grows and introduces challenging issue of RAG hallucination, primarily caused by the lack of corresponding parametric knowledge in LLMs. An efficient solution is to enhance the knowledge of LLMs at test-time. Parametric RAG (PRAG) addresses this by embedding document into LLMs parameters to perform test-time knowledge enhancement, effectively reducing inference costs through offline training. However, its high training and storage costs, along with limited generalization ability, significantly restrict its practical adoption. To address these challenges, we propose Dynamic Parametric RAG (DyPRAG), a novel framework that leverages a lightweight parameter translator model to efficiently convert documents into parametric knowledge. DyPRAG not only reduces inference, training, and storage costs but also dynamically generates parametric knowledge, seamlessly enhancing the knowledge of LLMs and resolving knowledge conflicts in a plug-and-play manner at test-time. Extensive experiments on multiple datasets demonstrate the effectiveness and generalization capabilities of DyPRAG, offering a powerful and practical RAG paradigm which enables superior knowledge fusion and mitigates RAG hallucination in real-world applications. Our code is available at https://github.com/Trae1ounG/DyPRAG.

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Forward citations

Cited by 18 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score.

  1. MetaSyn: A Benchmark for LLM Agents on Meta-Analysis Articles from Nature Portfolio

    cs.CL 2026-06 unverdicted novelty 8.0

    MetaSyn benchmark shows LLM pipelines recover at most 52.7% of ground-truth included studies due to screening failures on PI/ECO eligibility, despite 90.9% retrieval recall at K=200.

  2. MetaSyn: A Benchmark for LLM Agents on Meta-Analysis Articles from Nature Portfolio

    cs.CL 2026-06 unverdicted novelty 7.0

    LLM agents reach 90.9% retrieval recall at K=200 but recover at most 52.7% of ground-truth included studies because they cannot reliably apply PI/ECO eligibility criteria to topically similar distractors.

  3. MetaSyn: A Benchmark for LLM Agents on Meta-Analysis Articles from Nature Portfolio

    cs.CL 2026-06 conditional novelty 7.0

    MetaSyn is a stage-level benchmark of 442 meta-analyses showing LLM agents retrieve up to 90.9% of eligible studies but include at most 52.7% in their final reports.

  4. MetaSyn: A Benchmark for LLM Agents on Meta-Analysis Articles from Nature Portfolio

    cs.CL 2026-06 unverdicted novelty 7.0

    MetaSyn benchmark shows LLM agents recover at most 52.7% of relevant studies in meta-analysis pipelines due to failures in PI/ECO-based screening despite strong retrieval.

  5. RWGBench: Evaluating Scholarly Positioning in Related Work Generation

    cs.DL 2026-05 unverdicted novelty 7.0

    RWGBench is a citation-centric benchmark for related work generation built from 40k CS papers and a 100-paper test set, with multi-dimensional metrics that better match human expert judgment than standard similarity scores.

  6. Skill Retrieval Augmentation for Agentic AI

    cs.CL 2026-04 unverdicted novelty 7.0

    Introduces the SRA paradigm and SRA-Bench benchmark showing retrieval-based skill augmentation improves agent performance but skill incorporation remains a bottleneck regardless of retrieval quality.

  7. Skill Retrieval Augmentation for Agentic AI

    cs.CL 2026-04 unverdicted novelty 7.0

    Agents improve when they retrieve skills on demand from large corpora, yet current models cannot selectively decide when to load or ignore a retrieved skill.

  8. RWGBench: Evaluating Scholarly Positioning in Related Work Generation

    cs.DL 2026-05 accept novelty 6.5

    RWGBench evaluates related-work generation as citation decision-making (selection, placement, organization, discourse) rather than text similarity, exposing retrieval and generation failures that standard metrics miss.

  9. RWGBench: Evaluating Scholarly Positioning in Related Work Generation

    cs.DL 2026-05 conditional novelty 6.0

    RWGBench measures related-work generation by citation choices, and shows citation-focused metrics expose failures that text-similarity and LLM-judge scores miss.

  10. Enhancing Judgment Document Generation via Agentic Legal Information Collection and Rubric-Guided Optimization

    cs.CL 2026-05 unverdicted novelty 6.0

    Judge-R1 improves LLM judgment document generation by combining agentic legal information retrieval with GRPO-based rubric-guided optimization, outperforming baselines on the JuDGE benchmark.

  11. Skill Retrieval Augmentation for Agentic AI

    cs.CL 2026-04 unverdicted novelty 6.0

    Introduces SRA paradigm and SRA-Bench benchmark (5,400 tasks, 26,262 skills) showing retrieval improves performance but LLMs fail to selectively incorporate retrieved skills.

  12. FedMosaic: Federated Retrieval-Augmented Generation via Parametric Adapters

    cs.CL 2026-02 conditional novelty 6.0

    FedMosaic is a federated RAG system that encodes local documents as mask-gated LoRA adapters, clusters related documents into shared adapters, and selectively merges only relevant, low-conflict adapters at the server.

  13. MHA-RAG: Improving Efficiency, Accuracy, and Consistency by Encoding Exemplars as Soft Prompts

    cs.AI 2025-10 conditional novelty 6.0

    MHA-RAG encodes retrieved exemplars into order-invariant soft prompts via multi-head attention, claiming ~20-point effective-accuracy gains over RAG at ~10x lower inference FLOPs.

  14. RWGBench: Evaluating Scholarly Positioning in Related Work Generation

    cs.DL 2026-05 conditional novelty 5.5

    RWGBench evaluates related-work generation as citation decision-making with multi-dimensional metrics that track expert judgment better than ROUGE/BERTScore.

  15. Rethinking LoRA Memory Through the Lens of KV Cache Compression

    cs.CL 2026-06 unverdicted novelty 5.0

    Document LoRA acts as decoding-time parametric memory that recovers 13-21 ROUGE-L points under heavy KV cache compression in QA, performing best when the base model encodes the document and the adapter is used only at...

  16. How LoRA Remembers? A Parametric Memory Law for LLM Finetuning

    cs.CL 2026-05 unverdicted novelty 5.0

    Introduces Parametric Memory Law as power law for LoRA memory capacity and MemFT threshold-guided optimization for better memory fidelity.

  17. GRIP: In-Parameter Graph Reasoning through Fine-Tuning Large Language Models

    cs.CL 2025-11 reject novelty 5.0

    An LLM can memorize a knowledge graph into LoRA weights and answer relation/reasoning queries about it without graph context, but the evaluation partly trains on the test task.

  18. MemOS: A Memory OS for AI System

    cs.CL 2025-07 unverdicted novelty 5.0

    MemOS introduces a unified memory management framework for LLMs using MemCubes to handle and evolve different memory types for improved controllability and evolvability.