Providing Feedback

Providing Feedback

Feedback can be provided directly after each response. Click the thumbs up / down button to rate the response on a scale of 1–10 and add additional detail below.

  • If a response was amazing, we want to know exactly how it helped you and why.
  • If a response was subpar, we want to know too — with a clear understanding of why it did not meet the mark.

You can also flag issues in the SM Research Agent Slack help channel for triage.

Areas we would love feedback on

This is a non‑exhaustive list of metrics that help us improve the agent.

MetricDefinition
AccuracyAre the facts, numerical values, citations, and biological claims correct when compared with authoritative sources?
CompletenessDoes the answer cover all major aspects you would reasonably expect (e.g. pathways + cell types + methods), without obvious gaps?
ClarityIs the wording easy to follow — no ambiguous jargon, no convoluted sentences?
ConcisenessDoes the response avoid unnecessary detail, repetition, or filler while still conveying the essentials?
RelevanceHow well does the content stay on‑topic with the user’s explicit question and experiment context?
StructureIs information organized logically (e.g. bullet lists, headings, ordered steps)?
ToneIs the voice professional, collaborative, and appropriate for peer‑level scientific discussion (neither too casual nor too formal)?
UsefulnessWill the answer directly help you — e.g. design an experiment, interpret data, or save search time?
HallucinationDegree to which the response invents facts, citations, or statements not supported by sources or common biological knowledge.
FormattingTechnical presentation: code blocks render, figures load, citations link, and symbols (µM, ΔG) display correctly.
LatencyIs the time to receive responses for a specific query impacting how often you will use the agent?

Viewing previous chats

To access your chat history from a previous conversation, click the magnifying glass icon on the left‑hand side to open your previous chats.