# bytedance/pasa

PaSa -- an advanced paper search agent powered by large language models. It can autonomously make a series of decisions, including invoking search tools, reading papers, and selecting relevant references, to ultimately obtain comprehensive and accurate results for complex scholarly queries.

Repository: https://github.com/bytedance/pasa
Canonical: https://ross.abutalabs.com/products/pasa
Language: Python
License: Apache-2.0
License Family: permissive
Topics: research
Last push: 2025-05-27T11:09:11+00:00

## Health v2 (maintenance only)
Score: 31/100 (v2, computed 2026-09-02T17:46:02.011165+00:00)
- activity 23, release rhythm 35, longevity 44
- inputs: {"age_days": 618, "days_push": 463, "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 1649, forks 122 (observed 2026-08-28T04:05:16.758173+00:00)

## What it is
PaSa is an LLM-powered paper search agent from ByteDance that autonomously invokes search tools, reads papers, and selects relevant references to answer complex scholarly queries. It consists of two 7B agents (Crawler and Selector) trained with reinforcement learning on the synthetic AutoScholarQuery dataset, and is also available as a hosted demo.

## Use cases
- find papers for a detailed academic research query
- comprehensive literature search beyond Google Scholar
- search for papers cited within related work sections
- filter collected papers for relevance to my research question
- run an autonomous research assistant for scholarly queries
- evaluate LLM agents on academic search benchmarks

## When to choose
- you need comprehensive recall on complex or fine-grained scholarly queries
- you want an open-source, self-hostable alternative to Google Scholar or GPT-based search
- you need a research artifact with trained models and datasets available on Hugging Face

## When to avoid
- you need general web search rather than academic paper search
- you lack GPU resources to run 7B models locally and don't want to use the hosted demo
- you need a production-grade search service with SLAs

## Facets
- artifact type: application
- maturity: active
- function: agent-framework, search-engine, llm-inference, rag, machine-learning
- domain: artificial-intelligence, large-language-models
- platform: python, self-hosted
- tags: academic-paper-search, llm-agent, reinforcement-learning, scholarly-search, crawler-agent, acl-2025, ai-agents, search, natural-language-processing, linux, docker

## Member repositories
- bytedance/pasa (main) score 31

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:05:16.758173+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 2026-08-30T03:45:07.985596+00:00, confidence not recorded.
  - readme: https://github.com/bytedance/pasa (fetched 2026-08-28T04:05:16.758173+00:00, sha 551278318184)
- Data as of 2026-08-30T08:39:29.467469+00:00.