FractalAIResearch/Fathom-Synthesizer-4B
Fathom-Synthesizer-4B by FractalAIResearch is a 4 billion parameter model designed for open-ended synthesis and report generation within the Fathom-DeepResearch agentic system. It excels at transforming search and investigation traces into citation-dense reports, utilizing a novel plan-then-write protocol. With a context length of 40960 tokens, this model is optimized for long-horizon information retrieval and synthesis tasks.
Loading preview...
Fathom-Synthesizer-4B: DeepResearch Synthesis Model
Fathom-Synthesizer-4B, developed by FractalAIResearch, is a 4 billion parameter model specifically engineered for open-ended synthesis and report generation. It forms a core component of the Fathom-DeepResearch agentic system, which achieves state-of-the-art performance in open-weights categories for search-intensive benchmarks and outperforms closed-source agents on the DeepResearch-Bench synthesis benchmark.
Key Capabilities
- Open-Ended Synthesis: Designed to generate comprehensive, structured reports from complex information.
- Citation-Dense Reporting: Synthesizes findings into reports rich with citations, ensuring verifiability.
- Plan-Then-Write Protocol: Utilizes a novel protocol for report generation, involving question decomposition, evidence-to-section mapping, and insight planning.
- Long Context Handling: Supports a context length of 40960 tokens, enabling processing of extensive search traces.
- Integration with Fathom-Search-4B: Works in conjunction with Fathom-Search-4B, which is optimized for long-horizon, evidence-seeking through live web search.
Good For
- Automated Report Generation: Creating detailed reports from research data or search results.
- Information Synthesis: Consolidating disparate pieces of information into coherent narratives.
- Deep Research Applications: Powering agentic systems that require advanced information retrieval and structured output.
- Knowledge Management: Transforming raw data into organized, citable knowledge bases.