From Facts to Conclusions : Integrating Deductive Reasoning in Retrieval-Augmented LLMs

Fuente: arXiv
Salvato in:
Dettagli Bibliografici
Autori principali: Mishra, Shubham, Jain, Samyek, Mehrishi, Gorang, Tiwari, Shiv, Sharma, Harsh, Narang, Pratik, Kumar, Dhruv
Natura: Preprint
Pubblicazione: 2025
Soggetti:
Accesso online:
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866908720332537856
author Mishra, Shubham
Jain, Samyek
Mehrishi, Gorang
Tiwari, Shiv
Sharma, Harsh
Narang, Pratik
Kumar, Dhruv
author_facet Mishra, Shubham
Jain, Samyek
Mehrishi, Gorang
Tiwari, Shiv
Sharma, Harsh
Narang, Pratik
Kumar, Dhruv
contents Retrieval-Augmented Generation (RAG) grounds large language models (LLMs) in external evidence, but fails when retrieved sources conflict or contain outdated or subjective information. Prior work address these issues independently but lack unified reasoning supervision. We propose a reasoning-trace-augmented RAG framework that adds structured, interpretable reasoning across three stages : (1) document-level adjudication, (2) conflict analysis, and (3) grounded synthesis, producing citation-linked answers or justified refusals. A Conflict-Aware Trust-Score (CATS) pipeline is introduced which evaluates groundedness, factual correctness, refusal accuracy, and conflict-behavior alignment using an LLM-as-a-Judge. Our 539-query reasoning dataset and evaluation pipeline establish a foundation for conflict-aware, interpretable RAG systems. Experimental results demonstrate substantial gains over baselines, most notably with Qwen, where Supervised Fine-Tuning improved End-to-End answer correctness from 0.069 to 0.883 and behavioral adherence from 0.074 to 0.722.
format Preprint
id arxiv_https___arxiv_org_abs_2512_16795
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle From Facts to Conclusions : Integrating Deductive Reasoning in Retrieval-Augmented LLMs
Mishra, Shubham
Jain, Samyek
Mehrishi, Gorang
Tiwari, Shiv
Sharma, Harsh
Narang, Pratik
Kumar, Dhruv
Computation and Language
Artificial Intelligence
Computers and Society
Information Retrieval
Retrieval-Augmented Generation (RAG) grounds large language models (LLMs) in external evidence, but fails when retrieved sources conflict or contain outdated or subjective information. Prior work address these issues independently but lack unified reasoning supervision. We propose a reasoning-trace-augmented RAG framework that adds structured, interpretable reasoning across three stages : (1) document-level adjudication, (2) conflict analysis, and (3) grounded synthesis, producing citation-linked answers or justified refusals. A Conflict-Aware Trust-Score (CATS) pipeline is introduced which evaluates groundedness, factual correctness, refusal accuracy, and conflict-behavior alignment using an LLM-as-a-Judge. Our 539-query reasoning dataset and evaluation pipeline establish a foundation for conflict-aware, interpretable RAG systems. Experimental results demonstrate substantial gains over baselines, most notably with Qwen, where Supervised Fine-Tuning improved End-to-End answer correctness from 0.069 to 0.883 and behavioral adherence from 0.074 to 0.722.
title From Facts to Conclusions : Integrating Deductive Reasoning in Retrieval-Augmented LLMs
topic Computation and Language
Artificial Intelligence
Computers and Society
Information Retrieval
url https://arxiv.org/abs/2512.16795