Releases·Model Release
Open-sourcing AstaBrief, the fast report-generation model in Asta
Ai2 open-sourced AstaBrief 8B, an 8B model built on Qwen3-8B that turns a research question and retrieved literature excerpts into a cited scientific report. Trained with SFT and DPO on tens of thousands of real Asta queries, it ships today in Asta's Generate a report feature as Fast mode, generating reports ~3.5× faster than the Claude-powered Thinking mode. Weights and training data are open so institutions can run report generation on their own infrastructure.
0 engagement1 sourceFri, Oct 2, 2026, 03:19 PM
AI Overview
- Ai2 open-sources AstaBrief 8B, a Qwen3-8B fine-tune that turns a research question and retrieved literature excerpts into a cited report, with training data
- AstaBrief 8B powers Asta's Fast mode, averaging 51.1 seconds per report versus 178.5 seconds for Claude-powered Thinking mode, about 3.5x faster
- Trained with SFT and DPO rather than RL: 47K SFT examples from real Asta queries and 6K DPO pairs judged by GPT-4.1 and DeepSeek-R1 (95% human agreement)
- It generates the full report in one pass, bypassing the snippet summarization and clustering stages of the Claude-based pipeline
- Evaluated on SQABench-CS2 (200 CS questions) tracking rubric score, answer precision, and citation precision/recall, plus DeepScholarBench (63 queries)
- The source only notes that training and evaluation mostly ran in 2025 against then-frontier proprietary models; no full re-evaluation against today's frontier models
- Open weights let institutions run AstaBrief on their own infrastructure for sensitive or unpublished research; an example workflow adapts it to local PDFs
Related
Products·launched
open-academic-paper-gen: multi-agent pipeline generating draft papers with verified citations
An open-source multi-agent system that takes a research topic and produces a draft academic paper grounded in real literature. LLM agents handle drafting, while verification steps detect citation hallucinations and check whether claims are supported by sources. It targets researchers and writers who want LLM-assisted paper drafts without fabricated references.
An open-source multi-agent system that takes a research topic and produces a draft academic paper grounded in real literature. LLM agents handle drafting, while verification steps detect citation hallucinations and check whether claims are supported by sources. It targets researchers and writers who want LLM-assisted paper drafts without fabricated references.
31 engagement1 sourcegithubTue, Oct 6, 2026, 04:12 PM
Products·launched
Automated academic literature reviews using local LLMs via Ollama with verified verbatim quotes
An open-source tool that runs literature reviews locally through Ollama. It searches academic databases, retrieves full-text papers, and generates APA 7th-formatted reviews where the LLM is structurally prevented from fabricating quotes — each quote is verified verbatim against sources. It targets researchers who want accurate, citation-safe reviews without hallucinated references.
An open-source tool that runs literature reviews locally through Ollama. It searches academic databases, retrieves full-text papers, and generates APA 7th-formatted reviews where the LLM is structurally prevented from fabricating quotes — each quote is verified verbatim against sources. It targets researchers who want accurate, citation-safe reviews without hallucinated references.
26 engagement1 sourcegithubWed, Sep 2, 2026, 10:17 AM
Releases·Tool Release
EMBL AI Librarian: Life-Sciences Knowledge Layer for AI Agents
EMBL introduces AI Librarian, a knowledge layer that lets AI agents query life-sciences literature via natural language instead of complex syntax, returning evidence snippets rather than full papers. It addresses the growing need for agent-friendly access to resources like Europe PMC, which indexes over 40 million records.
EMBL introduces AI Librarian, a knowledge layer that lets AI agents query life-sciences literature via natural language instead of complex syntax, returning evidence snippets rather than full papers. It addresses the growing need for agent-friendly access to resources like Europe PMC, which indexes over 40 million records.
0 engagement1 sourcearxivThu, Jul 30, 2026, 02:00 PM
Releases·Tool Release
AI tool turns any research paper into an 'agent' that can collaborate and answer complex queries
Nature reports on an AI tool that converts any research paper into an interactive 'agent' able to collaborate and answer complex queries about the work. The approach points toward papers functioning as living interfaces rather than static documents. Details of the tool's architecture and availability are not included in the available excerpt.
Nature reports on an AI tool that converts any research paper into an interactive 'agent' able to collaborate and answer complex queries about the work. The approach points toward papers functioning as living interfaces rather than static documents. Details of the tool's architecture and availability are not included in the available excerpt.
1 engagement1 sourceredditThu, Sep 17, 2026, 12:56 AM
Social·Community
Community pushes back on conflating AI research agents with ChatGPT summaries
Practitioners argue that AI research agents must do more than summarize papers, such as decomposing vague goals into sub-questions and mapping citation networks. The discussion highlights a gap between current tools and the needs of real research workflows.
Practitioners argue that AI research agents must do more than summarize papers, such as decomposing vague goals into sub-questions and mapping citation networks. The discussion highlights a gap between current tools and the needs of real research workflows.
34 engagement1 sourceredditSat, Aug 22, 2026, 08:35 AM