International Journal For Multidisciplinary Research

E-ISSN: 2582-2160   •   Impact Factor: 9.24

A Widely Indexed Open Access Peer Reviewed Multidisciplinary Bi-monthly Scholarly International Journal

Call for Paper Volume 8, Issue 5 (September-October 2026) Submit your research before last 3 days of October to publish your research paper in the issue of September-October.
Submit Research Paper Join as a Reviewer

Plagiarism is checked by the leading plagiarism checker TurnItIn

Call for Paper

Volume 8 Issue 5
September-October 2026

Submit your research paper

Indexing Partners
Academia Advanced Sciences Index Bielefeld Academic Search Engine CiteSeer DRJI Google Scholar Independent Search Engine & Directory Network ISI (International Scientific Indexing) Issuu Mendeley Research Networks
RefSeek ResearcherId - Thomson Reuters ResearchGate Scirus Scribd Semantic Scholar UTeM - Universiti Teknikal Malaysia Melaka Wiki for Call for Papers WorldCat Zenodo

Evidence-Grounded Decision Support for IT Incident Triage with Local Language Models

Author(s) Dhananjay Ojha
Country India
Abstract IT incident triage involves determining if there is enough evidence for a safe fix, if details are missing and more information is needed, or if conflicting information is present and escalation is required. This research investigates the capabilities of the local language model, Ollama 3.2 3B, to perform this task using three methods: a prompt only baseline (G1), retrieval augmented generation (G2), and a structured output method (G3).
The development of a judgment only generation module means an easy to implement baseline. Using a 50 case development set, we got agreement rates of 72% for G1, 64% for G2, and 66% for G3. On a 30 case challenge set (G4), G1 got agreement of 76.7% vs. 40% and 50% for G2 and G3 respectively. The greatest variation was seen in detecting conflicts. The baseline G1 caught 9 out of 10 conflicts, while the other methods failed to catch any. We found G3 reliably enforced the structural schemas, but the relationships it identified were not sufficient to draw conflicts.
A judge-only human review of 90 outputs showed excellent consensus on judgment, truth and safety of the action suggested by the generated text. Human judges rated all outputs safe and judged them as extremely useful. Because of dataset and author conservatism, we believe enforcing a structural schema does not ensure correct relationships. Results reflect the performance in the environments tested.
Keywords incident management; retrieval-augmented generation; local language models; evidence grounding; decision support; synthetic evaluation
Field Computer > Artificial Intelligence / Simulation / Virtual Reality
Published In Volume 8, Issue 5, September-October 2026
Published On 2026-10-19
DOI https://doi.org/10.36948/ijfmr.2026.v08i05.90084
View / Download PDF File

Share this



E-ISSN 2582-2160
CrossRef

CrossRef DOI prefix of IJFMR is 10.36948/ijfmr

Downloads
Research Paper Format Copyright Permission Form and Undertaking Form

All research papers published on this website are licensed under Creative Commons Attribution-ShareAlike 4.0 International License, and all rights belong to their respective authors/researchers.

CC-BY-SA Open Access

About IJFMR
Fees & Payment
Current Issue
Publication Archive 		Submit Research Paper
Track Submission Status
Publication Guidelines
Publication Ethics
Peer Review & Plagiarism 		Join as a Reviewer
Editors & Reviewers
Reviewer Referral Program
Get Reviewer Membership Certi. 		Website/Journal Policies
Usage Policy
Content Policies
Privacy Policy
Contact Us +91-9687-828-838 editor@ijfmr.com
Powered by Sky Research Publication and Journals