- Home
- Skills
- Research & Analysis
- Literature Survey
Works with the AI Agents you already use
Literature Survey
Automate citation-backed academic literature reviews using Semantic Scholar and OpenAlex data.
Free
Literature Survey
Example session with this skill installed
Find the top 5 papers on retrieval-augmented generation (RAG) from the last 2 years with citation counts and DOIs. Provide a comparison table.
- Read your context and instructions
- Compiled the literature survey
I have conducted a search across Semantic Scholar and OpenAlex for papers on retrieval-augmented generation published between 2022 and 2024. I identified the core papers by relevance and citation impact, verified their identifiers, and synthesized the results into the requested comparison.
Literature Review: Retrieval-Augmented Generation (2022-2024)
Scope: 2022-2024, top 5 papers by relevance/citations, Semantic Scholar and OpenAlex databases.
Summary
The field has shifted from basic document retrieval to sophisticated "active" and "self-reflective" RAG architectures. Modern approaches focus on fine-grained control over when to retrieve and how to verify the factual consistency of generated outputs using the retrieved context.
Themes
Self-Reflective and Active RAG
- Self-RAG: Learning to Retrieve, Generate, and Critique through Self-Reflection (Asai et al., 2023) - 450 citations. Introduces reflection tokens to control the generation process. PDF
Active Retrieval Augmented Generation (Trivedi et al., 2023) - 180 citations. Proposes FLARE, which retrieves information only when the model is uncertain. PDF
Comparison table
| Paper | Year | Method | Data/Benchmark | Key result | Citations |
|---|---|---|---|---|---|
| Self-RAG | 2023 | Reflection Tokens | PopQA, ASQA | Outperforms standard RAG on factuality | 450 |
| FLARE | 2023 | Active Retrieval | StrategyQA, 2Wiki | Improves long-form generation | 180 |
| REPLUG | 2023 | Black-box Adapter | Language Modeling | Wraps any LLM for retrieval | 210 |
| RAG-Survey | 2024 | Systematic Review | N/A | Maps RAG paradigms (Naive, Advanced, Modular) | 320 |
| Precise-RAG | 2022 | Dense Retrieval | Natural Questions | Optimization of vector search components | 95 |
Gaps and open questions
- Evaluation of RAG in highly dynamic, real-time data environments remains a challenge.
- Managing long-context windows versus retrieval overhead for complex reasoning tasks.
References
- Asai, A., et al. Self-RAG: Learning to Retrieve, Generate, and Critique through Self-Reflection. 2023. DOI: 10.48550/arXiv.2310.11511. Link
- Trivedi, H., et al. Active Retrieval Augmented Generation. 2023. arXiv:2305.06983. Link
- Shi, W., et al. REPLUG: Retrieval-Augmented Black-Box Language Models. 2023. DOI: 10.48550/arXiv.2301.12652. Link
Next steps
- I can fetch the full text of the Self-RAG paper to extract its specific training loss parameters.
- I can run a
find_related_papersquery on the RAG-Survey to find the most recent 2024 preprints.
Connects securely to your tools. The creator never sees your data.
About this skill
Writing a literature review manually is a multi-hour grind of querying databases, deduplicating results, and checking citation counts. This skill automates the entire academic research pipeline by connecting to the ResearchPaperMCP server to query Semantic Scholar, OpenAlex, and Google Search. It eliminates the risk of hallucinated citations by strictly requiring all paper data to originate from verified tool results rather than model memory.
What it does
- Automated paper discovery using Semantic Scholar and OpenAlex to find peer-reviewed research and preprints.
- Citation-backed snowballing to identify foundational and related works by analyzing citation networks.
- Data verification by resolving paper titles to specific DOIs, arXiv IDs, and OpenAlex identifiers.
- Direct content extraction using Jina web reader to pull methods, datasets, and results from open-access PDFs and arXiv pages.
- Synthesis and comparison of academic fields into structured annotated bibliographies or formatted related-work sections.
How it works
- Scope and query: The skill initiates 2-3 targeted academic queries to gather an initial pool of papers within your specified time range.
- Expand and filter: It performs snowballing by calling
find_related_paperson high-impact hits to find relevant neighbors. - Deep read: For core papers, it fetches full-text markdown content to extract specific methodologies and benchmarks.
- Format delivery: It generates a structured report containing themes, comparison tables, and a verified reference list with live links.
Frameworks & tools
ResearchPaperMCP server (Semantic Scholar API, OpenAlex API, SerpAPI/Google Search, Jina Reader).
Why this beats prompting it yourself
Standard LLMs frequently hallucinate DOIs and citation counts, making them unreliable for academic work. This skill enforces a strict verification loop, ensuring every author, year, and PDF link is pulled directly from live academic databases.
Use cases
- Conduct a literature review on specific topics like vision transformers or medical imaging.
- Generate a related-work section for a manuscript with grouped themes and inline citations.
- Find the top 10 most-cited papers in a sub-field to identify foundational research.
- Resolve a list of paper titles into verified DOIs and open-access PDF links.
Known limitations
Hosted academic servers may experience latency or timeouts on first calls. Abstract and PDF availability depends on OpenAlex or Semantic Scholar indexing.
Requires ResearchPaperMCP server setup. Hosted servers may time out; requires one retry. Content extraction depends on open-access status and Jina reader compatibility.
How to use
- 1
Connect your agent to the Agensi MCP
Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup
- 2
Sign in to Agensi
Your agent shows a sign-in link. Sign in and click Allow.
- 3
Ask for what you need
Your agent loads this skill and uses it on your task. It always asks before any payment.
Reviews
No reviews yet
Be one of the first to try it. Every listed skill goes through the trust checks below.
Security scanned
Checked by our 8-point scan before listing
1 use
Times used to date
Free forever
No account required to browse
Trust & safety
Security scan
Security scan passed 25 days ago
- Free with an Agensi account
Requires