mattpocock / mattpocock/evalite

Feature: Allow scorer functions to return arrays of scores

Open
#304 1 comment 0 reactions 0 assignees View on GitHub

Nobody has claimed this yet.

enhancement
Dominant language
TypeScript
Stars
1.7k
Forks
102
PR merge metrics
No merged PRs in 30d

Description

Feature Request: Allow scorer functions to return arrays of scores

Problem

Currently, scorer functions can only return a single Score object. However, some use cases involve LLM graders that assess multiple criteria in a single call and return multiple pass/fail results.

Use Case

I have an LLM grader that performs a one-shot assessment of multiple criteria and returns a pass/fail for each criterion. For example, evaluating if a response is accurate, complete, and clear - all in a single LLM call for efficiency.

Currently, I need to either:

  • Make separate LLM calls for each criterion (inefficient and costly)
  • Write custom logic to split the results into multiple scorer calls
  • Use only a single aggregate score
Proposed Solution

Allow scorers to return either a single Score or an array of Score[], which would be flattened into the final scores array.

This appears to be achievable with a simple .flat() here: https://github.com/mattpocock/evalite/blob/e04b48ff260d8966e1fa3b2a6955c9abae9adcb9/packages/evalite/src/evalite.ts#L89 (plus type changes to allow Score | Score[]).

Contributor guide

Open the contributing guide

First steps

  1. Read the whole issue, then the project's contributing guide.
  2. Comment on the issue to say you are picking it up — it saves two people doing the same work.
  3. Fork the repository and make your change on a branch.
  4. Open a pull request that references the issue number.

Research direction

Start in packages/evalite/src/evalite.ts at the linked location around line 89, then trace the scorer result types. Update the accepted result shape and flatten returned score arrays into the final scores array; done means both single Score results and arrays of Score results are supported.

Written by the indexing model from the issue text.

Assessment

Tech stack
typescript
Domain
ai
Issue type
Feature
Difficulty
2/5
Estimated time
1-3 hours
Activity status
Stale
Clarity
Clearly specified
Newbie friendliness
55/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.