Deep Residual Learning for Image Recognition
222,082 citationsDeeper neural networks are more difficult to train. We present a residual learning framework to ease the training of networks that are substantially deeper than those used previously.
Multimodal AI Research Workspace
From Any Source to Research Papers, PDFs & Voice via RAG & MCP
Free. Bring your own key (BYOK).
The Rivul AI workspace. The sidebar lists Write: Papers, Research, Ask Rivul; Agents: Rivul MCP, Agents, Runs, Skills; Sources: Library, Search papers, Connectors; Convert: Document OCR, Transcribe, Listen; Tools: Citation tools. In the main view, Search papers (300M+ records), Document OCR (a PDF) and Transcribe (an interview) feed Agent run, which produces a cited draft whose numbered citations [1], [2] and [3] point to the paper page, the benchmark page and the interview time they came from.
Research Agents Grounded in Your Sources
Rivul AI (rivul.ai) is a free Multimodal AI Research Workspace for students, engineers, academics, researchers, and everyone who works from sources: turn PDFs, scans and recordings into text, ask your own documents, run AI agents, and search 300M+ academic papers.
Agent runs
Research agents search, read and write a cited report, and every run keeps its sources.
Grounded citations
Answers point to their sources: paper, page, passage.
Multimodal ingestion
OCR and Transcribe turn documents and recordings into searchable text.
In the workspace
Everything in the Rivul Workspace
- AgentsResearch agents that search papers and your library, then write a cited report on your own AI key.
- ResearchAsk a question across 300M+ papers and read an answer linked to its sources.
- Rivul MCPSearch the Rivul Index from Claude Code, Codex and Cursor, or from your own apps with the REST API.
- LibraryAdd papers by DOI, arXiv, .bib or Zotero, and keep up to 5 PDFs of 10 MB each.
- Search papers300M+ research records by title, author, topic or DOI. No AI key needed.
- ConnectorsBring in files from Google Drive and Google Docs, and your Zotero library.
- Document OCRPDFs, scans and images to text or Markdown: 10 files a batch, 25 MB and 50 pages each.
- TranscribeAudio, video and browser recordings to timestamped text: 10 files a batch, 500 MB each.
- Ask RivulAsk your library: answers from your own papers and PDFs, cited to the passage.
Also in Tools:Citation tools
Explore 300M+ Scholarly Works
Search needs no API key and no LLM: it runs on the Rivul Index.
Very Deep Convolutional Networks for Large-Scale Image Recognition
75,540 citationsIn this work we investigate the effect of the convolutional network depth on its accuracy in the large-scale image recognition setting.
Rethinking the Inception Architecture for Computer Vision
30,936 citationsConvolutional networks are at the core of most state of-the-art computer vision solutions for a wide variety of tasks.
Product tour
See It in Action
Search papers Search 315,518,142 indexed papers for “speculative decoding large language models”, with citation counts and open-access labels.
Real screens from the Rivul AI workspace, one per sidebar page.Meet your workspace
Free
Bring Your Own Key (BYOK)
- Free, no cardRivul AI is free for every account.
- Your own AI keyAI features run on your own key from Google Gemini, OpenAI, Anthropic, xAI, Mistral or Meta. The provider bills you directly.
- Rivul's own serversOCR, transcription and search run on Rivul's own servers, with no AI provider involved.
Frequently Asked Questions
What is Rivul AI?
Rivul AI (rivul.ai) is a free Multimodal AI Research Workspace for anyone who works from sources: students, engineers, academics, researchers, and everyday people. It brings academic papers from 300M+ indexed works, PDFs and scans, recordings, and your own library into one place, then works across them with Ask Rivul, Research, and AI agents.
What sources can Rivul AI work from?
Scholarly papers from the Rivul Index, PDFs you upload or add by DOI, arXiv, BibTeX, Google Drive, or Zotero, scanned documents and images through OCR, and audio or video through Transcribe.
Do I need an API key to search papers?
No. Search runs on the Rivul Index of 300M+ scholarly records and needs no key. Your own AI key is used only for AI features.
What do research agents do?
An agent has a goal, sources, and an output format. Give it a question and it searches papers and your library, reads the most relevant ones, and writes a report that cites each claim, using your own AI key. Every run is kept in Runs.
Can I use Rivul AI from Claude, Cursor, or Codex?
Yes. Rivul MCP is a remote MCP server with read-only tools for scholarly search and paper records. Every account can use it at no cost under fair-use rate limits.
Is my unpublished work private?
When you request an AI feature, the relevant text and passages are sent to the AI provider you connected, using your own API key. Rivul AI never trains models on your drafts or library and never sells your content.
How are bibliography entries created?
Reference records are resolved from scholarly metadata services such as Crossref and DataCite, then formatted by fixed rules. A resolved record does not prove that generated prose is correct or supported, so inspect the source before you rely on it.
Can Rivul AI turn scans and recordings into text?
Yes. Document OCR returns the text of PDFs, scans and images as plain text or Markdown, and Transcribe turns meetings, interviews, lectures and voice notes into timestamped text. Both run on Rivul AI's own servers and need no AI key.