TrustMargin: Training-Free Arbitration between Parametric Memory and Retrieved Evidence in Large Language Models

• URL: http://arxiv.org/abs/2606.08397v1
• Date: Sun, 07 Jun 2026 01:25:09 GMT
• Title: TrustMargin: Training-Free Arbitration between Parametric Memory and Retrieved Evidence in Large Language Models
• Abstract summary: We study this post-generation conflict as answer-level source arbitration.<n> given Direct and RAG answers from the same frozen model, decide which source to trust.<n>We propose TRUSTMARGIN, a training-free, plug-and-play arbitration layer that scores the two existing candidates with the model's own likelihoods.
• Score: 10.379427318055434
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: Large language models answer knowledge-intensive questions using both parametric memory and retrieved evidence, but neither source is uniformly reliable. Retrieval can fill knowledge gaps, yet distracting passages may override correct closed-book answers. We study this post-generation conflict as answer-level source arbitration: given Direct and RAG answers from the same frozen model, decide which source to trust. We propose TRUSTMARGIN, a training-free, plug-and-play arbitration layer that scores the two existing candidates with the model's own likelihoods. It combines a parametric-prior margin, which tests whether memory accepts the retrieved answer, with an evidence-binding margin, which discounts passage-only salience and measures question-specific support. TRUSTMARGIN selects between Direct and RAG without fine-tuning, external judges, or additional generation. Across 2WIKIMQA and CWQA with three LLaMA scales, TRUSTMARGIN consistently improves over Direct generation and BM25-RAG, recovers part of the Direct/RAG oracle gap, and generalizes to multiple training-free RAG pipelines.

Related papers

• BalanceRAG: Joint Risk Calibration for Cascaded Retrieval-Augmented Generation [8.129733777508434]
  Large language models (LLMs) can enhance factuality via retrieval-augmented generation (RAG)<n>Applying RAG to every query is unnecessary when the model-only answer is reliable.<n>We develop BalanceRAG to certify threshold pairs at a target risk level.
  arXiv  Detail & Related papers  (2026-05-19T16:38:55Z)
• To Fuse or to Drop? Dual-Path Learning for Resolving Modality Conflicts in Multimodal Emotion Recognition [87.22988227382329]
  Multimodal emotion recognition (MER) benefits from combining text, audio, and vision, yet standard fusion often fails when modalities conflict.<n>We propose Dual-Path Conflict Resolution (DCR), a unified framework that learns when to fuse and when to drop modalities.
  arXiv  Detail & Related papers  (2026-05-06T13:11:33Z)
• Reinforce to Learn, Elect to Reason: A Dual Paradigm for Video Reasoning [17.78591587115095]
  We introduce Reinforce to Learn, Elect to Reason (RLER), a dual paradigm that decouples learning to produce evidence from obtaining a reliable answer.<n>RLER achieves state of the art across all benchmarks and delivers an average improvement of 6.3% over base models.
  arXiv  Detail & Related papers  (2026-04-06T03:01:52Z)
• Probing the Trajectories of Reasoning Traces in Large Language Models [4.599673637363014]
  We propose a protocol to probe the trajectories of reasoning traces in large language models.<n>We find that accuracy and decision commitment consistently increase as the percentage of provided reasoning tokens grows.<n>We show that trajectory probing provides diagnostics for efficient and safer deployment of reasoning models.
  arXiv  Detail & Related papers  (2026-01-30T16:45:16Z)
• From Verifiable Dot to Reward Chain: Harnessing Verifiable Reference-based Rewards for Reinforcement Learning of Open-ended Generation [52.62655622099456]
  We propose reinforcement learning with verifiable reference-based rewards (RLVRR)<n>Instead of checking the final answer, RLVRR extracts an ordered linguistic signal from high-quality references (i.e., reward chain)<n>In this way, RLVRR decomposes rewards into two dimensions: content, which preserves deterministic core concepts, and style, which evaluates adherence to stylistic properties.
  arXiv  Detail & Related papers  (2026-01-26T14:39:58Z)
• Reference-Free Rating of LLM Responses via Latent Information [53.463883683503106]
  We study the common practice of asking a judge model to assign Likert-scale scores to free-text responses.<n>We then propose and evaluate Latent Judges, which derive scalar ratings from internal model signals.<n>Across a broad suite of pairwise and single-rating benchmarks, latent methods match or surpass standard prompting.
  arXiv  Detail & Related papers  (2025-09-29T12:15:52Z)
• SIRAG: Towards Stable and Interpretable RAG with A Process-Supervised Multi-Agent Framework [7.37561751991963]
  We propose a process-supervised multi-agent framework to bridge the gap between retriever and generator.<n>The proposed framework is modular and plug-and-play, requiring no modification to the retriever or generator.
  arXiv  Detail & Related papers  (2025-09-17T09:09:28Z)
• Question Decomposition for Retrieval-Augmented Generation [2.6409776648054764]
  We propose a RAG pipeline that incorporates question decomposition into sub-questions.<n>We show that question decomposition effectively assembles complementary documents, while reranking reduces noise.<n>Although reranking itself is standard, we show that pairing an off-the-shelf cross-encoder reranker with LLM-driven question decomposition bridges the retrieval gap on multi-hop questions.
  arXiv  Detail & Related papers  (2025-07-01T01:01:54Z)
• CoRe-MMRAG: Cross-Source Knowledge Reconciliation for Multimodal RAG [53.950029990391066]
  Cross-source knowledge textbfReconciliation for Multimodal RAG (CoRe-MMRAG)<n>We propose a novel end-to-end framework that effectively reconciles inconsistencies across knowledge sources.<n>Experiments on KB-VQA benchmarks show that CoRe-MMRAG achieves substantial improvements over baseline methods.
  arXiv  Detail & Related papers  (2025-06-03T07:32:40Z)
• R3-RAG: Learning Step-by-Step Reasoning and Retrieval for LLMs via Reinforcement Learning [60.17074283370798]
  Retrieval-Augmented Generation (RAG) integrates external knowledge with Large Language Models (LLMs) to enhance factual correctness and hallucination.<n>We propose $textbfR3-RAG$, which uses $textbfR$einforcement learning to make the LLM learn how to $textbfR$eason and $textbfR$etrieve step by step, thus retrieving comprehensive external knowledge and leading to correct answers.
  arXiv  Detail & Related papers  (2025-05-26T12:25:37Z)
• Retrieval-Augmented Generation with Conflicting Evidence [57.66282463340297]
  Large language model (LLM) agents are increasingly employing retrieval-augmented generation (RAG) to improve the factuality of their responses.<n>In practice, these systems often need to handle ambiguous user queries and potentially conflicting information from multiple sources.<n>We propose RAMDocs (Retrieval with Ambiguity and Misinformation in Documents), a new dataset that simulates complex and realistic scenarios for conflicting evidence for a user query.
  arXiv  Detail & Related papers  (2025-04-17T16:46:11Z)
• Merging Generated and Retrieved Knowledge for Open-Domain QA [72.42262579925911]
  COMBO is a compatibility-Oriented knowledge Merging for Better Open-domain QA framework. We show that COMBO outperforms competitive baselines on three out of four tested open-domain QA benchmarks.
  arXiv  Detail & Related papers  (2023-10-22T19:37:06Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.