Skip to content
System One

NandhaKishorM/laya

Decision model from Convai Innovations that fully fine-tunes an encoder and adds a typed-decision head: a 421M English checkpoint on ModernBERT-large and a 322M multilingual one on mmBERT-base, with a router picking between them by detected language. Returns choice, score and noul answers, and reports expected calibration error of 0.081 after temperature fitting, down from 0.466 as shipped.

laya-multilingual with max_len=8192: 16 to 18 of 20 requests correct with up to about 4,000 tokens of text before them, more variable beyond
Image from github.com

NandhaKishorM/laya on GitHub

Stars
21,445
Forks
1,828
Language
Python
License
Apache-2.0
Last push
24 Sep 2026

Read from GitHub on . Counts change daily.

How it uses Jev

 * call `ctx.skip(...)` to short-circuit inference, an end hook may rewrite the results.
 * See hooks.ts. `hooksRaise` overrides the Agent's setting for this call.
 */
async systemOne(
  state: unknown,
  questions: Record<string, QuestionDef>,
  opts: PredictOptions = {},
): Promise<SystemOneResult> {
  return (await this._predictHooked([state], questions, opts))[0];
}

View in laya-ts/src/agent.ts

Open on GitHub

PrimitivesChoiceScoreNoul

More like this

Keep browsing