View PDF ↗
PDF Viewer
Open Full PDF

BUILDER'S SANDBOX

Build This Paper

Use an AI coding agent to implement this research.

OpenAI Codex
OpenAI CodexAI Agent

Lightweight coding agent in your terminal.

Claude Code
Claude CodeAI Agent

Agentic coding tool for terminal workflows.

AntiGravity IDE
AntiGravity IDEScaffolding

AI agent mindset installer and workflow scaffolder.

Cursor
CursorIDE

AI-first code editor built on VS Code.

VS Code
VS CodeIDE

Free, open-source editor by Microsoft.

MVP Investment

$9K - $12K
6-10 weeks
Engineering
$8,000
Cloud Hosting
$240
SaaS Stack
$300
Domain & Legal
$100

6mo ROI

2-4x

3yr ROI

10-20x

Lightweight AI tools can reach profitability quickly. At $500/mo average contract, 20 customers = $10K MRR by 6mo, 200+ by 3yr.

Talent Scout

H

Hexi Jin

University of California, San Diego

S

Stephen Liu

University of California, San Diego

Y

Yuheng Li

University of California, San Diego

S

Simran Malik

University of California, San Diego

Find Similar Experts

AI experts on LinkedIn & GitHub

References (17)

[1]
OpenAI GPT-5 System Card
2025Aaditya K. Singh, Adam Fry et al.
[2]
DeepSeek-V3.2: Pushing the Frontier of Open Large Language Models
2025DeepSeek-AI, A. Liu et al.
[3]
TRUE: A Reproducible Framework for LLM-Driven Relevance Judgment in Information Retrieval
2025Mouly Dewan, Jiqun Liu et al.
[4]
LLM-based Relevance Assessment for Web-Scale Search Evaluation at Pinterest
2025Han Wang, Alex Whitworth et al.
[5]
Humanity's Last Exam
2025Long Phan, Alice Gatti et al.
[6]
Citations and Trust in LLM Generated Responses
2025Yifan Ding, M. Facciani et al.
[7]
Fine-grained Hallucination Detection and Mitigation in Language Model Mathematical Reasoning
2024Ruosen Li, Ziming Luo et al.
[8]
RAGProbe: An Automated Approach for Evaluating RAG Applications
2024Shangeetha Sivasothy, Scott Barnett et al.
[9]
DebateQA: Evaluating Question Answering on Debatable Knowledge
2024Rongwu Xu, Xuan Qi et al.
[10]
RAGBench: Explainable Benchmark for Retrieval-Augmented Generation Systems
2024Robert Friel, Masha Belyi et al.
[11]
Towards Understanding Factual Knowledge of Large Language Models
2024Xuming Hu, Junzhe Chen et al.
[12]
ARES: An Automated Evaluation Framework for Retrieval-Augmented Generation Systems
2023Jon Saad-Falcon, O. Khattab et al.
[13]
SWE-bench: Can Language Models Resolve Real-World GitHub Issues?
2023Carlos E. Jimenez, John Yang et al.
[14]
Dataset of Natural Language Queries for E-Commerce
2021A. Papenmeier, Dagmar Kern et al.
[15]
HotpotQA: A Dataset for Diverse, Explainable Multi-hop Question Answering
2018Zhilin Yang, Peng Qi et al.
[16]
Quora Question Pairs
2017Zihang Chen, Hongbo Zhang et al.
[17]
MS MARCO: A Human Generated MAchine Reading COmprehension Dataset
2016Daniel Fernando Campos, Tri Nguyen et al.

Founder's Pitch

"SourceBench enhances AI answer credibility by benchmarking web source quality."

AI BenchmarksScore: 5View PDF ↗

Commercial Viability Breakdown

0-10 scale

High Potential

2/4 signals

5

Quick Build

4/4 signals

10

Series A Potential

3/4 signals

7.5

Sources used for this analysis

arXiv Paper

Full-text PDF analysis of the research paper

GitHub Repository

Code availability, stars, and contributor activity

Citation Network

Semantic Scholar citations and co-citation patterns

Community Predictions

Crowd-sourced unicorn probability assessments

Analysis model: GPT-4o · Last scored: 2/18/2026

🔭 Research Neighborhood

Generating constellation...

~3-8 seconds

Why It Matters

As AI becomes an integral tool for information dissemination, the accuracy and quality of its outputs are crucial, especially in domains like finance, health, or law where decisions rely on credible sources.

Product Angle

Build a quality assurance tool that integrates with AI platforms to automatically assess and enhance the quality of web sources cited in AI answers.

Disruption

Could replace naive keyword-based source checking systems with a multi-faceted framework, raising the standard for source credibility in AI responses.

Product Opportunity

The tool can serve enterprises and legal professionals who need to ensure the reliability of AI-generated content. Market potential is significant given the growing use of AI in decision-making roles.

Use Case Idea

A browser extension using SourceBench algorithms to identify and label the quality of sources cited in AI-generated answers for professionals in research and legal fields.

Science

SourceBench evaluates web sources that LLMs cite by using an eight-metric framework that includes relevance, accuracy, objectivity, freshness, and clarity, ensuring holistic source quality rather than just textual relevance.

Method & Eval

The system evaluated 3996 web sources across 12 systems using human-aligned automated evaluation metrics in both content and meta-attributes, aiming for high correlation with manual scoring.

Caveats

The reliance on manual labeling for initial calibration could introduce bias, and there's potential difficulty in scaling across different domains or languages.

Author Intelligence

Hexi Jin

University of California, San Diego

Stephen Liu

University of California, San Diego

Yuheng Li

University of California, San Diego

Simran Malik

University of California, San Diego

Yiying Zhang

University of California, San Diego, GenseeAI Inc.
yiying@ucsd.edu