# GAIR-NLP/ASI-Arch

AlphaGo Moment for Model Architecture Discovery.

Repository: https://github.com/GAIR-NLP/ASI-Arch
Canonical: https://ross.abutalabs.com/products/asi-arch
Homepage: https://gair-nlp.github.io/ASI-Arch
Language: Python
License: Apache-2.0
License Family: permissive
Last push: 2025-12-03T07:41:28+00:00

## Health v2 (maintenance only)
Score: 43/100 (v2, computed 2026-09-02T17:46:02.011165+00:00)
- activity 55, release rhythm 35, longevity 29
- inputs: {"age_days": 411, "days_push": 273, "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 1183, forks 218 (observed 2026-08-28T04:03:54.433777+00:00)

## What it is
A multi-agent framework that lets an LLM autonomously conduct end-to-end research on neural network architecture discovery, specifically linear attention mechanisms. It includes a discovery pipeline, MongoDB-based architecture database, and cognition base, and ships 106 discovered state-of-the-art linear attention architectures.

## Use cases
- automate neural architecture discovery with LLM agents
- search for novel linear attention mechanisms
- run autonomous AI-driven scientific research pipelines
- store and retrieve experiment history for architecture search
- explore a gallery of discovered attention architectures

## When to choose
- you want an autonomous multi-agent system for architecture research
- you need a starting set of state-of-the-art linear attention designs
- you want to reproduce or extend LLM-driven scientific discovery experiments

## When to avoid
- you need a production-ready attention implementation library
- you lack GPU resources for large-scale empirical validation
- you want a simple hyperparameter search tool rather than architecture discovery

## Facets
- artifact type: framework
- maturity: active
- function: agent-framework, llm-inference, machine-learning, database, workflow-automation
- domain: artificial-intelligence, large-language-models, deep-learning
- platform: python
- tags: autonomous-research, linear-attention, multi-agent, scientific-discovery, architecture-search, ai-agents, research, docker, linux

## Member repositories
- GAIR-NLP/ASI-Arch (main) score 43

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:03:54.433777+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 2026-08-30T06:24:30.664810+00:00, confidence not recorded.
  - readme: https://github.com/GAIR-NLP/ASI-Arch (fetched 2026-08-28T04:03:54.433777+00:00, sha dcd5cc4bcbf6)
  - homepage: https://gair-nlp.github.io/ASI-Arch (fetched 2026-08-29T12:31:43.513209+00:00, sha ed691f71df84)
- Data as of 2026-08-30T08:39:29.467469+00:00.