PaperQA2, Agentic RAG for Science
We're excited to release PaperQA2, an open source RAG library specialized to work with the scientific literature. We've seen some really compelling results with it (https://paper.wikicrow.ai), like superhuman performance at question answering and summarization when compared with expert scientists. PaperQA2 is a major overhaul of our prior PaperQA system, it includes automatically obtained rich metadata for each paper, a CLI to work with local papers directly, a local full-text search engine for keywords searches over PDF files, a state-of-the-art algorithm for LLM-based re-ranking…
What it does
In the maker’s words, at launch
We're excited to release PaperQA2, an open source RAG library specialized to work with the scientific literature. We've seen some really compelling results with it (https://paper.wikicrow.ai), like superhuman performance at question answering and summarization when compared with expert scientists. PaperQA2 is a major overhaul of our prior PaperQA system, it includes automatically obtained rich metadata for each paper, a CLI to work with local papers directly, a local full-text search engine for keywords searches over PDF files, a state-of-the-art algorithm for LLM-based re-ranking and contextual summarization (RCS), the ability to agentic RAG, and support for all LiteLLM compatible LLMs. We hope that folks are able to use PaperQA2 to help improve their scientific workflows and the way they interact with the literature. It's enabled many cool projects for us like (WikiCrow and ContraCrow), and we hope to see many more in the future.
Does the same job
all alternatives →- ALA labelling tool to easily extract and label Wikipedia data2021 · ▲116
Hi HN! I am Maria, solo founder of DataQA (https://dataqa.ai/), a tool to search and label documents for various NLP tasks (e.g. entity extraction, entity linking, etc). I have worked as a data scientist and ML engineer for the better part of a decade, and over that time have specialised mainly in applications involving natural language processing (NLP). One of the key questions I have always had at the back of my mind is whether my time was well spent. Whenever I spent more time on feature engineering or trying different models, I always wondered whether I would get better…
- OSOpen-Source Notebooks for Building Agentic RAG Architectures2025 · github.com · ▲6
Hey HN! We’ve published a series of open-source notebooks showcasing Advanced RAG and Agentic architectures, and we’re excited to share our latest compilation of Agentic RAG techniques! These Colab-ready notebooks are designed to be plug-and-play, making it easy to integrate them into your projects. We're actively expanding the repository and would love your input to shape its future. What Advanced RAG technique should we add next? Drop your ideas in the comments or open an issue on GitHub!
- HKHybrid Knowledge Graph and RAG for Legal Documents (Learning Project)2025 · github.com · ▲8
Built this as a toy project to understand knowledge graphs by tackling a real problem: traditional RAG fails badly on legal documents because it misses interconnections between sections. The system actually combines both approaches on every query - gets semantic matches via TF-IDF, retrieves structural relationships from Neo4j, then feeds both contexts to OpenAI for comprehensive answers. Used the Indian Income Tax Act as test data since legal documents have natural graph structures. Queries like "What sections reference Section 80C?" get both the reference network AND content explanations.…
- IRInteractive research papers (a big step up from ArXiv HTML)Nov 2025 · sciencestack.ai · ▲9
I’ve been frustrated with PDFs and found arXiv HTML lacking, so I built a fully interactive paper reader. Features: • Hover references, citations, equations • Light/dark mode • Auto-generated dependency graphs for definitions/lemmas/theorems • Table of contents that syncs with scroll • Highlighting + annotations • “Copy raw LaTeX” anywhere Featured paper: Video models are zero-shot learners and reasoners (Veo 3) https://www.sciencestack.ai/arxiv/2509.20328v2
More ai this month
the category →
I trained a 125M-parameter transformer to autocomplete piano performances in real time (~108 notes/sec on an iPhone 15). The idea is basically GitHub Copilot or Tabnine, except instead of prompting it with code, you prompt it by playing a few notes on a MIDI piano. The model then continues what you played, entirely on-device. The app is free if anyone wants to try it. Happy to answer questions about the model, training, Core ML, or the many things that didn't work.
AI · 17d ago · simedw.com
Astute▲585Automate your B2B brand going viral, with new media creators
AI · 18d ago · company-app.joinastute.com
Hey HN, Henry from Cactus here! We previously released Cactus Needle, a 14MB agentic LLM for tool call, device use, and structured extraction for phones, wearables, smart homes, small robots and microcontrollers. We got really great feedback here, and have now incorporated the suggestions to release Needle 2. The whole model is a single 14MB binary that runs a full session in 28MB of RAM; 45m parameters at 2bit compression. Needle hits 500 tokens/sec decode speed on a Raspberry Pi 5, sits between 400-1,500 tokens/sec on VR devices like Meta Quest 3S and Apple Vision Pro, and ranges…
AI · 26d ago · cactuscompute.com
Launched alongside, September 2024
the whole month →
BeforeSunset AI 2.0▲1,267Personalized AI daily planning that suits your life
AI · 2024 · beforesunset.ai
