Article Per Year (5 Year)
p-Index From 2021 - 2026
0.444
P-Index
This Author published in this journals
All Journal Journal of Technology Informatics and Engineering
Wenhao Su
Computer Science, UCSD, CA, USA
Computer Science & IT

Published : 2 Documents Claim Missing Document
Claim Missing Document
Check
Articles

Found 2 Documents
Search

 1 
Budgeted Multi-Hop Retrieval Agent for Compositional Question Answering: A Retrieval-Policy Evaluation on the Official MultiHop-RAG Benchmark Wenhao Su; Siyu Chen; Chloe Zhao
Journal of Technology Informatics and Engineering Vol. 4 No. 3 (2025): DECEMBER | JTIE : Journal of Technology Informatics and Engineering
Publisher : University of Science and Computer Technology

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.51903/jtie.v4i3.543

Abstract

Multi-hop question answering requires a retrieval system to assemble several complementary evidence documents before an answer module can reason reliably. Single-shot retrieval is efficient, but it often misses later-hop evidence when a question combines source, time, comparison, and entity constraints. This paper evaluates a budgeted multi-hop retrieval agent for compositional question answering on the official MultiHop-RAG benchmark. The benchmark contains 2,556 queries and 609 news-article corpus documents, with answerable evidence distributed across two to four documents. Four retrieval policies are compared under the same sparse lexical scorer: fixed top-k retrieval, iterative retrieval, query decomposition, and the proposed budgeted retrieval agent. The revised evaluation frames the task as retrieval-policy evaluation rather than as a full free-form generative QA system: retrieval-conditioned EM/F1 are reported together with evidence recall, MRR, retrieval rounds, selected documents, and context-token cost. On the official data, the budgeted agent achieves the strongest overall retrieval-conditioned EM/F1 at 62.75% and the highest final evidence recall at 74.67%, using 3.011 average retrieval calls and 509.7 average context tokens. Query decomposition improves over fixed top-k and iterative retrieval but is less stable across question types. Fixed top-k is cheapest but incomplete on longer chains. The four-hop results remain difficult for every policy, showing that a fixed 620-token controller should be extended with hop-aware or dynamic budget allocation. The findings support a moderated contribution claim: explicit budget control is useful for auditable multi-hop retrieval, but it should be evaluated as a cost-accuracy trade-off rather than as a universally dominant RAG architecture.
Narrative-Aware Scientific Claim Verification Agent with Evidence Ranking for ClimateCheck Wenhao Su; Siyu Chen; Ethan Qian
Journal of Technology Informatics and Engineering Vol. 5 No. 1 (2026): APRIL | JTIE : Journal of Technology Informatics and Engineering
Publisher : University of Science and Computer Technology

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.51903/jtie.v5i1.549

Abstract

Climate misinformation often combines a factual proposition with a recognizable narrative, such as denying observed warming, rejecting human causation, minimizing impacts, attacking mitigation, or casting doubt on climate science. This paper presents a lightweight narrative-aware scientific claim verification agent for the official ClimateCheck setting. The revised evaluation uses the official annotated ClimateCheck training data, the official publications corpus of 394,269 abstracts, and a claim-level validation split of the annotated data. The public ClimateCheck test file is treated as a blind claim list because its public fields do not contain verification or narrative labels. The system combines hashed BM25, TF-IDF retrieval, latent semantic analysis, narrative-family probabilities, and a logistic-regression verifier. Full-corpus retrieval shows that BM25 remains the strongest first-stage retriever, with Recall@10 = 0.466, while the narrative-aware hybrid obtains Recall@10 = 0.444. In the judged candidate reranking setting, the narrative-aware ranker obtains the highest Candidate Recall@1 = 0.789 and MAP = 0.848, compared with 0.759 and 0.843 for TF-IDF. End-to-end verification remains difficult: the BM25 top-1 pipeline reaches Macro-F1 = 0.408, while the narrative-aware pipeline reaches Macro-F1 = 0.355. Claim-level narrative evaluation no longer produces a perfect score; single-label top-family Macro-F1 is 0.422, and fine-grained multi-label CARDS-code Macro-F1 is 0.098. These results show that narrative information is useful for reranking already plausible evidence candidates, but it does not replace strong lexical retrieval and does not by itself solve claim verification.
Co-Authors Chloe Zhao Ethan Qian Siyu Chen Siyu Chen