Adds GSM-PLUS

[NathanHB]

Results for HuggingFaceTB/SmolLM2-1.7B-Instruct

uv run lighteval vllm "model_name=HuggingFaceTB/SmolLM2-1.7B-Instruct"  "lighteval|gsm_plus|0|0"   --use-chat-template

Task	Version	Metric	Value		Stderr
all		extractive_match	0.213	±	0.0043
lighteval:gsm_plus:0	0	extractive_match	0.213	±	0.0043

[Copilot]

Pull Request Overview

Adds support for a new GSM-Plus task by registering its configuration and prompt handler.

• Introduce a gsm_plus task in default_tasks.py
• Implement gsm_plus prompt logic in default_prompts.py

Reviewed Changes

Copilot reviewed 2 out of 2 changed files in this pull request and generated 1 comment.

File	Description
src/lighteval/tasks/default_tasks.py	Register new gsm_plus task configuration
src/lighteval/tasks/default_prompts.py	Add prompt function for filtering and formatting

Comments suppressed due to low confidence (1)

src/lighteval/tasks/default_tasks.py:7963

• Add tests to cover the new gsm_plus task configuration (e.g., prompt generation and evaluation flow) to ensure it behaves as expected.

gsm_plus = LightevalTaskConfig(

[HuggingFaceDocBuilderDev]

The docs for this PR live here. All of your documentation changes will be reflected on that endpoint. The docs are available until 30 days after the last update.

[lewtun]

Really nice eval! Before merging, could you run 1-2 models from their table to see if we get similar results?