## Flow Diagram: Paper Selection Process ### Overview The image is a flow diagram illustrating the process of selecting papers based on specific criteria. It starts with an initial search and progresses through stages of identification, screening, eligibility, and inclusion, with the number of papers decreasing at each stage. The diagram also outlines inclusion and exclusion criteria. ### Components/Axes * **Initial Search:** "Initial Search with search keyword and IC1" * Flows into four databases: Clarivate Web of Science, OpenReview.net, arXiv, and Scopus. * Clarivate Web of Science: 1235 papers * OpenReview.net: ~1000 papers * arXiv: 9180 papers * Scopus: 3270 papers * **Identification:** * Label: i * Text: "Total Paper" * Value: 14685 * **Screening:** * Label: ii * Text: "Run EC1, EC2, EC3" * Value: 2932 * **Eligibility:** * Label: iii * Text: "Run IC2, IC3, IC4" * Value: 182 * **Inclusion:** * Label: iv * Text: "Run IC5" * Value: 64 * **Inclusion Criteria (IC):** * IC1: "Between year January 1, 2021 and September 15, 2025. Must be in Scopus, Web of Science, OpenReview, ArXiv" * IC2: "Published in A*/A/B/Notable Conferences and Q1/Q2 Journals" * IC3: "Preprint although published by high ranked universities researchers from reputable labs and/or domain experts" * IC4: "Scanning full texts" * IC5: "Studies must address fact-checking, misinformation detection, or hallucination mitigation in LLMs" * **Exclusion Criteria (EC):** * EC1: "Duplicate items" * EC2: "Studies publish in other language than English" * EC3: "Scan irrelevant titles and abstracts" ### Detailed Analysis or ### Content Details The diagram starts with an initial search, which yields different numbers of papers from four databases: Clarivate Web of Science (1235), OpenReview.net (~1000), arXiv (9180), and Scopus (3270). These papers are then combined, resulting in a total of 14685 papers at the "Identification" stage. The papers then undergo "Screening" based on exclusion criteria EC1, EC2, and EC3, reducing the number to 2932. Next, "Eligibility" is assessed using inclusion criteria IC2, IC3, and IC4, further reducing the number to 182. Finally, "Inclusion" is determined by applying inclusion criterion IC5, resulting in a final selection of 64 papers. ### Key Observations * The number of papers decreases significantly at each stage of the selection process. * The initial search yields the most papers from arXiv (9180), followed by Scopus (3270). * The exclusion criteria (EC1, EC2, EC3) have a substantial impact on reducing the number of papers during the screening stage. * The inclusion criteria (IC2, IC3, IC4, IC5) further refine the selection process, leading to a final set of 64 papers. ### Interpretation The flow diagram illustrates a systematic approach to paper selection, starting with a broad search and narrowing down the results based on specific inclusion and exclusion criteria. The significant reduction in the number of papers at each stage highlights the importance of these criteria in identifying relevant and high-quality studies. The diagram also demonstrates the relative contribution of different databases to the initial pool of papers. The process ensures that the final selection of papers meets the defined requirements for the research or analysis being conducted.

\n ## Flowchart: Systematic Review Search and Selection Process ### Overview This flowchart illustrates the process of a systematic review, detailing the number of records identified, screened, assessed for eligibility, and ultimately included in the review. The process begins with an initial search across four databases and progresses through stages of identification, screening, eligibility, and inclusion, with associated record counts at each step. Inclusion/Exclusion Criteria (IC1-IC5) are defined on the right side of the image. ### Components/Axes The flowchart consists of boxes representing stages in the review process, connected by arrows indicating the flow of records. Circles represent the initial databases searched. The right side of the image contains a legend defining the Inclusion/Exclusion Criteria (IC1-IC5). * **Initial Search Databases:** Clarivate Web of Science, OpenReview.net, arXiv, Scopus * **Stages:** Identification, Screening, Eligibility, Inclusion * **Inclusion/Exclusion Criteria (IC1-IC5):** * IC1: BETWEEN YEAR JANUARY 1, 2021 AND SEPTEMBER 15, 2025. MUST BE IN SCOPUS, WEB OF SCIENCE, OPENREVIEW, ARXIV * IC2: PUBLISHED IN A/A*/B/NOTABLE CONFERENCES AND Q1/Q2 JOURNALS * IC3: PREPRINT ALTHOUGH PUBLISHED BY HIGH RANKED UNIVERSITIES RESEARCHERS FROM REPUTABLE LABS AND/OR DOMAIN EXPERTS * IC4: SCANNING FULL TEXTS * IC5: STUDIES MUST ADDRESS FACT-CHECKING, MISINFORMATION DETECTION, OR HALLUCINATION MITIGATION IN LLMS * **Additional Labels:** * i: Total Paper * ii: Run EC1, EC2, EC3 * iii: Run IC2, IC3, IC4 * iv: Run IC5 ### Detailed Analysis or Content Details The flowchart begins with an initial search across four databases: 1. **Clarivate Web of Science:** 1235 records 2. **OpenReview.net:** Approximately 1000 records (~1000) 3. **arXiv:** 9180 records 4. **Scopus:** 3270 records These records are combined, resulting in a total of 14685 records at the **Identification** stage (i). * **Screening (ii):** 2932 records remain after applying EC1, EC2, and EC3. * **Eligibility (iii):** 182 records are deemed eligible after applying IC2, IC3, and IC4. * **Inclusion (iv):** 64 records are ultimately included after applying IC5. The circles representing the databases are numbered 1 through 4, corresponding to the order listed above. The arrows connecting the databases to the "Identification" box are dashed, indicating the initial search and record retrieval. The arrows connecting the stages are solid, indicating the flow of records through the review process. ### Key Observations The number of records decreases significantly at each stage of the review process. The largest reduction occurs between the Identification and Screening stages (14685 to 2932), suggesting that the initial screening criteria (EC1, EC2, EC3) are highly selective. The final inclusion of only 64 records from an initial search of over 14,000 indicates a rigorous and focused review process. ### Interpretation This flowchart demonstrates a typical systematic review methodology. The substantial reduction in records at each stage highlights the importance of well-defined inclusion and exclusion criteria in ensuring the quality and relevance of the included studies. The specific criteria (IC1-IC5) indicate that this review focuses on recent (2021-2025) research related to fact-checking, misinformation detection, and hallucination mitigation in Large Language Models (LLMs), published in reputable sources (Scopus, Web of Science, OpenReview, arXiv, A/A*/B journals, Q1/Q2 conferences). The process is designed to filter out irrelevant or low-quality studies, resulting in a small, highly relevant set of studies for synthesis and analysis. The use of approximate values (e.g., ~1000) suggests a degree of uncertainty in the initial record counts, potentially due to variations in search strategies or database coverage.

## [Process Diagram]: Systematic Literature Review Screening Funnel ### Overview This image is a flowchart illustrating the multi-stage screening process for a systematic literature review. It details the initial search across four academic databases, followed by a four-phase filtering funnel (Identification, Screening, Eligibility, Inclusion) that progressively reduces the number of candidate papers based on specific inclusion (IC) and exclusion (EC) criteria. The final output is a set of 64 included studies. ### Components/Axes The diagram is organized into three main sections: 1. **Top Row (Database Search):** Four rounded rectangles represent the initial data sources, each with a logo, name, and the number of retrieved records. 2. **Left Column (Screening Funnel):** A vertical flowchart showing the four phases of the review process. Each phase is labeled (i, ii, iii, iv), has a title (e.g., IDENTIFICATION), describes the action taken (e.g., "Run EC1, EC2, EC3"), and shows the resulting number of papers in an oval. 3. **Right Column (Criteria Legend):** A vertical list of criteria, each in a colored oval connected to a dashed circle label (IC1-IC5, EC1-EC3). Green ovals denote Inclusion Criteria (IC), and pink ovals denote Exclusion Criteria (EC). **Detailed Component Transcription:** * **Database Search (Top Row, Left to Right):** * **Initial Search:** "Initial Search with search keyword and IC1" (in a dark blue oval). * **Clarivate Web of Science™:** Logo + "Clarivate Web of Science™", Number: **1235**. Connected to step **1**. * **OpenReview.net:** Logo + "OpenReview.net", Number: **~1000**. Connected to step **2**. * **arXiv:** Logo + "arXiv", Number: **9180**. Connected to step **3**. * **Scopus:** Logo + "Scopus", Number: **3270**. Connected to step **4**. * **Screening Funnel (Left Column, Top to Bottom):** * **Phase i - IDENTIFICATION:** "Total Paper" -> Arrow to oval with number **14685**. * **Phase ii - SCREENING:** "Run EC1, EC2, EC3" -> Arrow to oval with number **2932**. * **Phase iii - ELIGIBILITY:** "Run IC2, IC3, IC4" -> Arrow to oval with number **182**. * **Phase iv - INCLUSION:** "Run IC5" -> Arrow to oval with number **64**. * **Criteria Legend (Right Column, Top to Bottom):** * **IC1 (Green):** "BETWEEN YEAR JANUARY 1, 2021 AND SEPTEMBER 15, 2025" / "MUST BE IN SCOPUS, WEB OF SCIENCE, OPENREVIEW, ARXIV" * **IC2 (Green):** "PUBLISHED IN A*/A/B/NOTABLE CONFERENCES AND Q1/Q2 JOURNALS" * **IC3 (Green):** "PREPRINT ALTHOUGH PUBLISHED BY HIGH RANKED UNIVERSITIES RESEARCHERS FROM REPUTABLE LABS AND/OR DOMAIN EXPERTS" * **IC4 (Green):** "SCANNING FULL TEXTS" * **IC5 (Green):** "STUDIES MUST ADDRESS FACT-CHECKING, MISINFORMATION DETECTION, OR HALLUCINATION MITIGATION IN LLMS" * **EC1 (Pink):** "DUPLICATE ITEMS" * **EC2 (Pink):** "STUDIES PUBLISH IN OTHER LANGUAGE THAN ENGLISH" * **EC3 (Pink):** "SCAN IRRELEVANT TITLES AND ABSTRACTS" ### Detailed Analysis The process begins with an initial search using a keyword and Inclusion Criterion 1 (IC1), which defines the date range (Jan 1, 2021 - Sep 15, 2025) and the four source databases. This yields a combined total of **14,685** papers (1235 + ~1000 + 9180 + 3270). The funnel then applies filters in sequence: 1. **Screening Phase:** Exclusion Criteria EC1 (duplicates), EC2 (non-English), and EC3 (irrelevant titles/abstracts) are applied. This reduces the pool from **14,685** to **2,932** papers, a reduction of ~80%. 2. **Eligibility Phase:** Inclusion Criteria IC2 (venue quality), IC3 (author/lab credibility), and IC4 (full-text scan) are applied. This further reduces the count from **2,932** to **182** papers, a reduction of ~94%. 3. **Inclusion Phase:** The final Inclusion Criterion IC5 (specific topic relevance to fact-checking/misinformation/hallucination in LLMs) is applied. This results in the final set of **64** included papers, a reduction of ~65% from the previous stage. ### Key Observations * **Massive Initial Retrieval:** The arXiv database contributed the largest share of initial records (9180, ~62% of the total). * **Steep Filtering Curve:** The most significant reduction occurs between the Screening and Eligibility phases (from 2932 to 182), indicating that venue quality, author credibility, and full-text relevance are highly selective filters. * **Final Yield:** The process results in a final inclusion of **64** papers, representing approximately **0.44%** of the initial search results. * **Criteria Specificity:** The inclusion criteria progress from broad (date/source) to highly specific (topic relevance to LLM hallucination mitigation). The exclusion criteria target common methodological issues (duplicates, language, irrelevance). ### Interpretation This diagram transparently documents a rigorous, multi-stage systematic review methodology. The funnel visualization effectively communicates the selectivity of the process. The sharp declines in paper counts at each stage demonstrate the application of increasingly stringent quality and relevance filters. The criteria reveal the study's precise focus: recent (2021-2025), high-impact research (A*/A/B conferences, Q1/Q2 journals) from reputable sources, specifically addressing the critical issue of factual reliability in Large Language Models (fact-checking, misinformation, hallucination mitigation). The final set of 64 papers likely represents the core, high-quality literature on this specific topic within the defined timeframe. The process is designed to be reproducible, with each filtering step explicitly linked to defined criteria (IC/EC).

## Flowchart: Research Study Selection Process for LLM Hallucination Mitigation ### Overview The flowchart illustrates a multi-stage research study selection process for evaluating hallucination mitigation techniques in Large Language Models (LLMs). It begins with an initial search across four academic databases, followed by systematic screening, eligibility assessment, and final inclusion criteria application. The right panel details specific inclusion criteria (IC1-IC5) with color-coded annotations. ### Components/Axes **Left Panel (Flowchart):** 1. **Initial Search (Stage 1):** - Databases: Clarivate Web of Science (1235), OpenReview.net (~1000), arXiv (9180), Scopus (3270) - Total papers identified: 14,685 2. **Screening (Stage 2):** - Exclusion Criteria (EC1-EC3): Duplicate items, non-English publications, irrelevant titles/abstracts - Papers after screening: 2,932 3. **Eligibility (Stage 3):** - Inclusion Criteria (IC2-IC4): Published in A*/A/B journals/conferences, preprints from reputable institutions, full-text scanning - Papers after eligibility: 182 4. **Inclusion (Stage 4):** - Inclusion Criterion (IC5): Studies addressing fact-checking, misinformation detection, or hallucination mitigation in LLMs - Final included studies: 64 **Right Panel (Inclusion Criteria):** - **IC1:** Publication date between January 1, 2021, and September 15, 2025; must be in Scopus, Web of Science, OpenReview, or arXiv - **IC2:** Published in A*/A/B notable conferences or Q1/Q2 journals - **IC3:** Preprints from high-ranked universities/researchers in reputable labs/domain experts - **IC4:** Full-text scanning requirement - **IC5:** Focus on fact-checking, misinformation detection, or hallucination mitigation in LLMs ### Detailed Analysis - **Initial Search:** - arXiv contributes the most papers (9180), followed by Scopus (3270), Web of Science (1235), and OpenReview.net (~1000) - Total papers: 14,685 (sum of all database contributions) - **Screening Process:** - EC1 (duplicates), EC2 (non-English), EC3 (irrelevant titles/abstracts) reduce the pool from 14,685 to 2,932 - Color coding: Pink for screening criteria - **Eligibility Assessment:** - IC2 (publication venue), IC3 (institution reputation), IC4 (full-text access) filter to 182 papers - Color coding: Green for eligibility criteria - **Final Inclusion:** - IC5 (LLM hallucination focus) selects 64 studies - Color coding: Light blue for inclusion criteria ### Key Observations 1. **Publication Venue Concentration:** - arXiv dominates initial search (62.5% of total papers) - Scopus and Web of Science together represent 22.3% 2. **Screening Efficiency:** - 80.3% of papers eliminated during screening (11,753 removed) 3. **Eligibility Stringency:** - Only 12.4% of screened papers (182/1468) meet eligibility criteria 4. **Final Selection:** - 35.2% of eligible papers (64/182) satisfy IC5 requirements ### Interpretation This flowchart demonstrates a rigorous, multi-phase selection process typical of systematic reviews in computational linguistics. The hierarchical filtering: 1. **Broadens scope** through multi-database searches 2. **Applies venue-based filters** (IC1-IC4) to ensure academic rigor 3. **Focuses on technical relevance** through IC5's LLM-specific requirements The color-coded annotations (pink/green/blue) spatially ground the criteria application across stages. Notably, the 64 final studies represent 0.44% of the initial search corpus, highlighting the challenges in LLM research reproducibility and the need for standardized evaluation frameworks. The emphasis on preprints (IC3) and full-text access (IC4) suggests an attempt to balance academic rigor with practical accessibility in this rapidly evolving field.