> For the complete documentation index, see [llms.txt](https://docs.nested-knowledge.com/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.nested-knowledge.com/ai-transparency/model-cards/synthesis-exports.md).
# Model Cards: Synthesis/Exports
AI tools at this stage draft written summaries, structured abstracts, and conversational answers from a nest's live, extracted data.
| Tool | What it does | Model card |
| --------------------------------- | ------------------------------------------------------------------------------------------- | ---------- |
| [Smart Insights](#smart-insights) | Drafts evidence-backed written summaries from your extracted data | Live |
| [Smart Abstract](#smart-abstract) | Drafts a structured Methods, Results, and Discussion abstract from your nest's data | Live |
| [Synthesis Chat](#synthesis-chat) | Answers plain-language questions and generates tables/figures using your review's live data | Live |
Smart Insights
### Model Name: Smart Insights
**Version: 1.1**
### Overview
Smart Insights summarize tag extractions within a nest. They operate by:
* Selecting tags to summarize, automatically via relevance to the Research Question or from researcher input.
* For each tag:
* Grouping extractions to common Claims that address the tag
* Producing a narrative tying together the independent Claims as an answer to the tag
All 3 phases are powered by an LLM, GPT 5.6 Luna.
The purpose of the Extraction > Claim > Insight hierarchy is to contain LLM-driven responses to only the evidence extracted within a review. It also guarantees that **every** extraction in a review is either positively or negatively included in the Insight, minimizing and making transparent any bias from omission.
### Intended Use
#### Primary Purpose
\
To summarize compiled evidence acquired in a systematic review rapidly, comprehensively, and without extrapolation.
#### Intended Users
Smart Insights are intended for use and consumption by both authors and readers of SLRs:
* Systematic review teams
* Clinical researchers
* Evidence synthesis professionals
* Health technology assessment groups
### Evaluation
Smart Insights lack a quantitative benchmark, and instead have been qualitatively evaluated by Nested Knowledge research staff.
### Limitations
* Smart Insights only generate for individual tags (extractions) and do not synthesize combinations of tags
* Smart Insights are qualitative and do not generate quantitative, inferential summaries
* Smart Insights have only been tested to a few hundred extractions; the performance of scaling qualitative summaries beyond this amount is unclear.
Smart Abstract
### Model Name: Smart Abstract
**Version: 1.0**
### Overview
Smart Abstract summarize the methods employed and results discovered within a Nest. To accomplish this, it:
* Retrieves nest configurations & workflow settings
* Retrieves and summarizes Search, Tagging, and Critical Appraisal data within the nest
* Retrieves and summarizes Insights within the Nest
* Places all of this information into a conditional, templated Abtract format.
Summary and selection steps within this process are automated by an LLM, GPT 5.6 Luna.
### Intended Use
#### Primary Purpose
Summarize the methods employed and results discovered within a Nest so that a researcher not familiar with a review may quickly understand its findings and how they were produced.
#### Intended Users
Smart Abstract is intended for use and consumption by both authors and readers of SLRs:
* Systematic review teams
* Clinical researchers
* Evidence synthesis professionals
* Health technology assessment groups
Especially researchers who need a rapid summary of how a review was conducted and what is contained within it.
### Evaluation
Smart Abstract lacks a quantitative benchmark, and instead has been qualitatively evaluated by Nested Knowledge research staff. As a largely templated format on top of previously generated data, its output has been evaluated to be stable.
### Limitations
* Smart Abstract, by design, does not produce an Introduction section that would be familiar to SLR readers
* Smart Abstract only includes Insights in its Results section. Other forms of synthesis (e.g. Quantitative Insights, analyses derived by review authors) will not be generated, and the limitations applying to Smart Insights apply here.
Synthesis Chat
### Model Name: Synthesis Chat
**Version: 0.2**
### Overview
Synthesis Chat provides an exploratory chat interface to a review. It is structured as:
* An LLM agent, which the user interacts with
* Subagents, who may conduct tasks requested by the agent
* Predefined tool calls which retrieve & analyze data contained within a review. For example, PRISMA counts, search sizes and their import histories
Agents and subagents are powered by a variety of OpenAI models (main: GPT 5.6 series), with the specific model conditionally selected based on context.
The key constraint given to the agent is to limit responses to only contain information contained within the review. To this end it produces a full citation list, reported to the user, supporting each chat response.
### Intended Use
#### Primary Purpose
Provide an exploratory interface to methods and data extracted in a review, producing custom, interpretable answers and reports for review consumers.
#### Intended Users
Synthesis Chat is intended for use and consumption by both authors and readers of SLRs:
* Systematic review teams
* Clinical researchers
* Evidence synthesis professionals
* Health technology assessment groups
### Evaluation
A Synthesis Chat benchmark dataset is forthcoming. Synthesis Chat's quality has been evaluated internally by the Nested Knowledge research team for acceptability, but no formal reports on performance are offered at this time. Users should exercise caution in interpreting its outputs.
### Limitations
* Synthesis Chat tool calls may be insufficient to retrieve **all** data needed to answer a researcher inquiry, be it by volume of data or coverage of data types. This may produce incomplete answers.
* Synthesis Chat cannot answer questions that were not addressed by searched records & extractions within a review, by design. It will indicate this to the user, but by design, it will not take any action to "go back" or rework earlier parts of the review. Researchers must undertake this step in AutoLit by themselves.
* Evaluation of Synthesis Chat performance has not been formalized or reported.
---
# Agent Instructions
This documentation is published with GitBook. GitBook is the documentation platform designed so that both humans and AI agents can read, navigate, and reason over technical content effectively. Learn more at gitbook.com.
## Querying This Documentation
If you need additional information that is not directly available in this page, you can query the documentation by asking a question.
Perform an HTTP GET request on the following URL with the `ask` and `goal` query parameters:
```
GET https://docs.nested-knowledge.com/ai-transparency/model-cards/synthesis-exports.md?ask=&goal=
```
`ask` is the immediate question: it should be specific, self-contained, and written in natural language.
`goal` is what the user is ultimately trying to achieve, the reason they need the answer. Sharing it helps GitBook give you a better, more relevant answer. A goal is most helpful when it describes the outcome the user wants rather than restating the question. For example, with `ask=how do I create an API token`, a goal like `automate deployments from our CI pipeline` lets GitBook tailor the answer to that use case.
The response will contain a direct answer to the question and relevant excerpts and sources from the documentation.
Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.