# EleutherAI/lm-evaluation-harness

A framework for few-shot evaluation of language models.

Repository: https://github.com/EleutherAI/lm-evaluation-harness
Canonical: https://ross.abutalabs.com/products/lm-evaluation-harness
Homepage: https://www.eleuther.ai
Language: Python
License: MIT
License Family: permissive
Topics: evaluation-framework, language-model, transformer
Last push: 2026-08-26T16:38:13+00:00

## Health v2 (maintenance only)
Score: 89/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 99, release rhythm 71, longevity 100
- inputs: {"age_days": 2197, "days_push": 7, "days_rel": 114, "gap_med": 54.0, "n_releases_24m": 11}
- flags: none
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 13803, forks 3518 (observed 2026-08-28T04:11:04.825556+00:00)

## What it is
A Python framework for few-shot evaluation of language models across hundreds of standard benchmarks (HellaSwag, MMLU, BIG-Bench, etc.). It supports HuggingFace, vLLM, SGLang, and API-hosted models with configurable tasks via YAML and a refactored CLI.

## Use cases
- evaluate an LLM on standard benchmarks like MMLU or HellaSwag
- compare language models on few-shot tasks
- run the Open LLM Leaderboard task suite
- benchmark a HuggingFace or vLLM model
- evaluate models served via OpenAI-compatible APIs
- create custom evaluation tasks with YAML configs
- measure chain-of-thought reasoning performance

## When to choose
- you need reproducible, standardized LLM benchmark scores
- you want to evaluate models across many backends (HF, vLLM, SGLang, APIs)
- you need configurable few-shot prompting and custom tasks
- you're submitting or reproducing Open LLM Leaderboard results

## When to avoid
- you need broad multimodal (text+image) evaluation coverage - consider lmms-eval
- you want training or fine-tuning rather than evaluation
- you need a hosted no-code evaluation service

## Facets
- artifact type: framework
- maturity: active
- function: machine-learning, llm-inference, benchmarking, testing, cli
- domain: large-language-models, machine-learning, developer-tools
- platform: python, cli, windows
- tags: llm-evaluation, benchmarks, few-shot, huggingface, vllm, open-llm-leaderboard, model-evaluation, natural-language-processing, gpu, linux, macos

## Member repositories
- EleutherAI/lm-evaluation-harness (main) score 89

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:11:04.825556+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-29T17:13:01.956119+00:00, confidence not recorded.
  - readme: https://github.com/EleutherAI/lm-evaluation-harness (fetched 2026-08-28T04:11:04.825556+00:00, sha d2260704980a)
  - homepage: https://www.eleuther.ai (fetched 2026-08-29T08:07:19.363916+00:00, sha 46c7f1a08777)
  - site_page: https://www.eleuther.ai/about (fetched 2026-08-29T08:07:19.377512+00:00, sha 8a099f02f9a1)
  - site_page: https://www.eleuther.ai/releases (fetched 2026-08-29T08:07:19.373346+00:00, sha 6b771c6c6df0)
- Data as of 2026-08-30T08:39:29.467469+00:00.