I was hoping for this to announce a tool for research.

Anyone know of the best way to do something like:

"Find most relevant papers related to topic XYZ, download them, extract metadata, generate big-picture summary and entity-relationship graph"?

Having a nice workflow for this would be the best thing since sliced bread for hobbyists interested in niche science topics.

Recently found https://minicule.com which is free and lets you search + import, but it focuses more on "concept-extraction" than LLM synthesis/summary.

A while ago, I started working on two R packages for creating 'living reviews': metawoRld and DataFindR, see https://andjar.github.io/metawoRld/articles/conceptual_overv... . You do the broad literature search yourself, but the idea is to use LLMs to select relevant studies and perform data extraction in a structured, reproducible manner. The extracted data is stored in a git repository for collaboration and version tracking, with automated validation and website generation for presenting results.

"Structured and Reproducable"

PaperAI is also an option if you prefer open-source: https://github.com/neuml/paperai

Disclaimer: I'm the primary author of this project.

Check out https://elicit.com/

Seems potentially useful, thanks! Only drawback I can see is the small number of papers provided by the free plan, but that's reasonable I suppose.

I built a public literature review search tool for some graduate student friends that became pretty popular in the Santa Barbara area. It actually does exactly what you are describing.

It’s not neural network based: it leverages hierarchical mixture models to give a statistical overview of the data. It lets you build these analysis graphs via search or citation networks.

Example: https://platform.sturdystatistics.com/deepdive?search_type=e...

This is genuinely incredible, tried it using a recent-ish paper on the pharmacology and mechanisms of the Androgen Receptor and my mind is blown:

https://platform.sturdystatistics.com/deepdive?fast=1&q=http...

I've been trying to tackle this exact problem. Current process is to use exa.ai to collect a wide breadth of research papers. Do a summarization pass and convert to markdown. Search for more specific terms then give the relevant papers/context to Gemini 2.5 pro and say give me a summary. Looking for very specific resources and to be honest it's been a terrible process :|

Linking to a nearby thread in case this is helpful: https://news.ycombinator.com/item?id=44457928

My site, https://www.emergentmind.com, is exactly for this. It surfaces trending AI/ML/CS papers, summarizes them, links to social commentary, lets you read and download papers, links to topics, and more. Would love any feedback you have!

I’ve found a lot of success with https://www.undermind.ai/ though I’m not sure it has the graph you’re looking for

This also looks excellent, thank you!

Hi, I'm the creator of https://tatevlab.com. It does something similar + aiming to be something like a "spotify" for research papers (currently working on a feature to allow creating and sharing personal collections). It summarizes papers based on practical potential and you can find papers based on similarity. Feedback is welcome.

Their Chemistry LLM that's an iteration of ChemCrow is really useful, thank you!

Connectedpapers.com

emergentmind is pretty good

From the title, I had thought that this would be a new tool for searching science, such as searching the arxiv. But this is actually a survey.

I quote the conclusion of the survey:

---

In conclusion, rapid advancements in artificial intelligence, particularly large language models like OpenAI-o1 and DeepSeek-R1, have demonstrated substantial potential in areas such as logical reasoning and experimental coding. These developments have sparked increasing interest in applying AI to scientific research. However, despite the growing potential of AI in this domain, there is a lack of comprehensive surveys that consolidate current knowledge, hindering further progress. This paper addresses this gap by providing a detailed survey and unified framework for AI4Research. Our contributions include a systematic taxonomy for classifying AI4Research tasks, identification of key research gaps and future directions, and a compilation of open-source resources to support the community. We believe this work will enhance our understanding of AI’s role in research and serve as a catalyst for future advancements in the field.

---

I jumped at this because I'm a mathematician who has been complaining about the lack of effective mathematical search for several years.

How do you view o3? I personally find it superior to google search almost always. Do you find that it often misses key references? (also mathematician)

Google is completely inadequate at mathematical search. But here is a concrete problem that no search seems to handle: given some complicated integral (say, some contour integral involving a K-Bessel function), find where it appears in the literature.

Most search will totally fail, because this is made of math symbols. Embedding-based search will give various related things involving, say, integrals and Bessel functions. But then I end up opening Gradshteyn and Ryzhik and trying to find where in this book the relevant terrible integrals appear.

This is a common experience for analytic number theorists. And it's a lousy experience.

Have you found https://sugaku.net/ useful? It’s focused on math research

This paper is more of a meta-level overview than a hands-on solution

AI for Scientific Search yes. LLM for Scientific Search I am not sure. AI is not equivalent with LLM. I dislike it when people do it.

AI will have a brand crisis once LLMs get abandoned and researchers need to explain the public that the new AI (not LLM based) is different than the old AI (LLM based) which is different from the old AI (GOFAI)

> once LLMs get abandoned

See, you start making a good point in your rant, but then go too much and stop making sense. LLMs are not going to be abandoned. They've "solved" intent from natural language. They're here to stay.

Of course "AI" will get new things. And architectures might improve. And new things will be discovered and added to the tool box. But having the ability to use natural language as input is so invaluable that there's no way we'll just abandon it...

We will abandon it when we find something better. That is the lifecycle of technology.

I like zotero, I started vibe coding some integration for my workflow, the project is a bit clunky to build and iterate the development specially with gemini & claude. But I think that is the direction to take instead of reinvent from scratch something

I've been thinking about a plugin that auto-suggests related papers as I write

AI getting into scientific research is definitely impressive. But the more we use it, the more it feels like we're slowly getting too lazy to think on our own. Human judgment and intuition seem to be fading bit by bit.

"AI" is also the opposite of scientific research: word-suggestion algorithm which guess what is the most probable next part given a set of inputs. In the end, you'll still need to prove that your theory is right.

Always worth noting where the authors are affiliated and I don't remember ever hearing of bytedance breaking new ground in chemical or materials research so I'm sceptical about reading this...

Is there any intership opportunity for me

I wonder how well these models will hold up in messy, interdisciplinary real-world projects

Was expecting a product I can try out. But still, not disappointed.