Agentic Capabilities: Asta agents are built for scientific workflows. They provide source-cited results to reduce issues like hallucinations in research. Ai2 also released AstaBench to evaluate scientific AI agents. Asta's "Find Papers" feature delivers an LLM-powered search experience that reformulates search queries (or, prompts), follows citations, and explains the relevance of retrieved papers. It's been described as "Google Scholar on steroids" and a powerful tool for researchers navigating large volumes of scientific literature.
Conventional AI models are designed to perform specific tasks, but lack autonomy to adapt to complex scenarios. Agentic AI seeks to solve this problem. Agentic AI is a category of AI capable of independently making decisions, interacting with its environment, and optimizing processes without direct human intervention. Agentic AI represents a big advance vs. traditional AI by incorporating features such as self-learning, real-time adaptability, and multi-agent collaboration. Agentic AI still needs to be verified for scientific topics.
Agentic AI refers to advanced artificial intelligence (AI) agents to make decisions, plan, and execute actions to achieve specific objectives with minimal human intervention. AI agents can be used to perform small tedious tasks and actions on behalf of the user; they can integrate a spectrum of AI capabilities including interacting with websites and performing actions on behalf of the user.
Three components in ASTA
Asta's ecosystem brings together three parts to advance scientific AI; 1) Asta agents are tools to assist (not replace) human researchers performing complex tasks. To promote transparency, 2) AstaBench provides benchmarking framework for evaluating and comparing any agent—not just Asta. 3) Asta resources offers a set of software components and standards to help build, test, and refine scientific agents.
1) Asta Agents https://allenai.org/asta: the agent is an AI research assistant for scientists, helping with tasks e.g., literature reviews, evidence synthesis, and data analysis (in beta). AAs are transparent, citing sources for outputs and aim to support researchers by handling complex, multi-step tasks while maintaining scientific rigour. Features include finding relevant papers, summarizing literature with real citations, and analyzing datasets using natural language queries.
2) AstaBench https://allenai.org/asta/bench: a benchmarking framework with over 2,400 problems across 11 benchmarks in four categories (literature understanding, code execution, data analysis, and end-to-end discovery); evaluates AI agents on real-world scientific tasks, offering leaderboards to compare performance and cost efficiency. Asta’s agent scored 52.5% in initial evaluations, outperforming others such as GPT-5 mini and Claude 3.5 Haiku.
3) Asta Resources https://allenai.org/asta/resources: developer toolkit includes open-source agents, APIs, post-trained language models and a Scientific Corpus Tool built on Ai2’s Semantic Scholar (200M+ papers); supports building and testing scientific AI agents via Model Context Protocol (MCP).
Marketing
Asta is "built" for science, emphasizing verifiability and openness, and used by researchers at 194 institutions such as University of Chicago and University of Washington. Its goal is to accelerate tasks like therapeutic target identification. Future plans include experiment replication, hypothesis generation, and scientific programming. Ai2 invites scientific and AI communities to contribute to set a standard for trustworthy AI in research.
"...Asta, an integrated ecosystem with an agentic research assistant, [is] a benchmarking framework for AI agents, and developer resources.... our bold initiative to accelerate science by building trustworthy and capable agentic assistants for researchers, along the first comprehensive benchmarking system to bring clarity to the landscape of scientific AI. As AI use expands across the sciences, researchers in every field need helpful systems they can understand, verify, and trust. Asta is designed to fill this need.
Asta is an evolving initiative ...by releasing each component ...we will continue to improve scientific AI agents through real-world usage and feedback from the research and AI development communities. Realizing the potential of AI for science is a shared effort, and we invite you to learn more about what we're releasing, where we're headed, and how we can move forward together."
Presentation
Note: This presentation was selected by a librarian due to the presenter and their understanding of the product. As this is a marketing video, the claims of the video should be tested and verified.
Librarian criticism
Bottom line: For health sciences librarians, many of the AI tools are similar, and use Semantic Scholar as their search corpus. I'll be testing this new AI-powered search tool in the coming weeks but suspect it's similar to the others such as Elicit.com and Undermind.ai. It's unfortunate that the developer hasn't reached out to me as a librarian assessing AI-powered tools.
"...AI Agents [are] modular systems driven by Large Language Models (LLMs) and Large Image Models (LIMs) for narrow, task-specific automation. Generative AI is positioned as a precursor, with AI Agents advancing through tool integration, prompt engineering, and reasoning enhancements. In contrast, Agentic AI systems represent a paradigmatic shift marked by multi-agent collaboration, dynamic task decomposition, persistent memory, and orchestrated autonomy."
Disclaimer
Note: Please use your critical reading skills while reading entries. No warranties, implied or actual, are granted for any health or medical search or AI information obtained while using these pages. Check with your librarian for more contextual, accurate information.
