TEST RECORD · ← all field reports

FIELD REPORTS/FIELD NOTE/ISSUE #113

Answers yes/no and multiple-choice questions about any text, with a confidence score on every answer

Most of the time you do not need a model to write anything. You need it to answer a short question about a piece of text, and to tell you how sure it is. Laya does only that, and it does it in about the time it takes to blink.

MODELLaya
PUBLISHEDSeptember 24, 2026
READ TIME3 min
TESTED BYNeural Expedition
CATEGORYFIELD NOTE

Field notes

01What it does

You give Laya two things: a piece of text and a list of questions. Each question has a fixed shape. A yes or no question ("Does the customer threaten to cancel?"), a choice between options you name ("Which team should handle this: billing, technical, or sales?"), or a score on a scale you describe ("How urgent is this, from not urgent to blocking?"). It answers all of them in one pass, roughly 30 to 40 milliseconds on a modest GPU, and every answer comes with a probability. If it says billing at 0.94, that number is meant to be trusted: the model was trained so that reporting an honest probability is the only way to score well.

That is the part that makes it different from asking a chat model. Laya never generates text. There is no reply to parse, no made-up detail to check, and no waiting for tokens. The answer space is whatever you define in the question, so you can add a new label or a new question without retraining anything.

It comes in three checkpoints under one repo. The English one is the default. A multilingual one covers more than 100 languages and reads up to about 8,000 tokens of text, so a long document works too. A third is fine-tuned for four business workflows (invoices, security incidents, customer service, agent logs) and shows how much accuracy jumps once you train it on decisions from your own domain. A built-in router picks the right checkpoint by detecting the language of the input.

02How to try it

Open the Laya demo Space. Paste a real email or ticket into the state box, write two or three questions of different types, and run it. Read the probabilities next to each answer, not just the answer. Then change one word in the text, say turn "please refund" into "maybe refund", and run again to watch the confidence move. To run it yourself, install the laya package with pip and call the router; the checkpoints download on first use and run on CPU as well, just slower.

03Caveat

Out of the box the base checkpoints are only useful for simple, well-described questions; on the authors' own business benchmark they score near chance until fine-tuned, and they get weak once a single question has more than about 20 options. Treat it as a fast base to specialise, and check the confidence numbers on your own data before you trust them.

04What you can do with it

  • Triage a support inbox by department, urgency, and churn risk with one call.
  • Add a cheap guardrail in front of a chat model: is this request in scope, is it safe, does it need a human.
  • Route customer messages in any language to the right queue without translating them first.
  • Score agent logs or incident reports for review priority.
  • Tag a large pile of documents with your own labels, quickly enough to do it live.

Try the demo

View model page

Read this issue on neuralexpedition.com →