# eureka-research/Eureka

Official Repository for "Eureka: Human-Level Reward Design via Coding Large Language Models" (ICLR 2024)

Repository: https://github.com/eureka-research/Eureka
Canonical: https://ross.abutalabs.com/products/eureka-research-eureka
Homepage: https://eureka-research.github.io/
Language: Jupyter Notebook
License: MIT
License Family: permissive
Last push: 2024-05-03T07:31:13+00:00

## Health v2 (maintenance only)
Score: 27/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 0, release rhythm 35, longevity 76
- inputs: {"age_days": 1073, "days_push": 852, "days_rel": null, "gap_med": null, "n_releases_24m": 0}
- flags: no_releases
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 3189, forks 300 (observed 2026-08-28T04:07:48.095498+00:00)

## What it is
Eureka is a research library from NVIDIA that uses coding LLMs like GPT-4 to automatically generate reward functions for reinforcement learning via evolutionary in-context optimization. It outperforms human-engineered rewards across 29 RL environments and enables gradient-free RLHF.

## Use cases
- generate reward functions for RL training automatically
- train dexterous robot manipulation skills like pen spinning
- apply LLMs to reinforcement learning reward design
- implement gradient-free RLHF with human feedback
- run evolutionary optimization over reward code with GPT-4

## When to choose
- you need reward functions for IsaacGym robotics RL environments
- you want to research LLM-driven reward design or RLHF
- you lack expert reward engineering for complex manipulation tasks

## When to avoid
- you need a production system rather than research code
- your RL setup is not based on IsaacGym or the bundled environments
- you cannot access GPT-4-class LLM APIs

## Facets
- artifact type: library
- maturity: maintenance
- function: machine-learning, llm-inference, agent-framework, simulation
- domain: reinforcement-learning, robotics, large-language-models, artificial-intelligence
- platform: python
- tags: reward-design, reinforcement-learning, llm-code-generation, rlhf, isaacgym, research-code, pytorch, research, linux, gpu

## Member repositories
- eureka-research/Eureka (main) score 27

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:07:48.095498+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 2026-08-30T07:24:56.999413+00:00, confidence not recorded.
  - readme: https://github.com/eureka-research/Eureka (fetched 2026-08-28T04:07:48.095498+00:00, sha ff71e74aad1c)
  - homepage: https://eureka-research.github.io/ (fetched 2026-08-29T09:38:44.607189+00:00, sha b2b3c872df02)
- Data as of 2026-08-30T08:39:29.467469+00:00.