说说人话 BenchHUMAN COMMUNICATION LAB Research preview · v0.5
LANGUAGE IN CONTEXT, MEASURED BY PEOPLE

Good writing belongs
to a real situation.

A benchmark for the messages people would actually use.
Compare responses in context. Preserve disagreement. Show uncertainty.

Loading study status…

From a scenario to evidence

One transparent research workflow
01 / DESIGN

Freeze the study

Declare scenarios, systems and analysis choices before collecting judgments.

Versioned inputs
02 / OBSERVE

Let people compare

Blind identities, balance positions and record what each person would send or revise.

A · Tie · B
03 / UNDERSTAND

Inspect the evidence

See uncertainty, coverage and disagreement alongside the estimates.

No universal score
CHALLENGE COLLECTION

The hard part is often one word.

“Scheduled” is different from “confirmed.” An update is not a delivery guarantee. A warm reply should not invent a fact.

Commitment strengthPrivacy boundariesAudience adaptationUncertaintyScope precisionSocial context

The repository includes 72 Chinese scenario candidates: the 24-case pilot plus 48 new AI-authored challenges awaiting human review.

YOUR OWN STUDY

Bring an annotation packet.

Open a JSON packet supplied by your study coordinator. Progress stays in this browser until you export it.

Use assigned pseudonyms. The packet contains anonymous response IDs.