# salesforce/decaNLP

The Natural Language Decathlon: A Multitask Challenge for NLP

Repository: https://github.com/salesforce/decaNLP
Canonical: https://ross.abutalabs.com/products/decanlp
Language: Python
License: BSD-3-Clause
License Family: permissive
Topics: deep-learning, natural-language-processing, multitask-learning
Archived: true
Last push: 2025-05-01T17:29:03+00:00

## Health v2 (maintenance only)
Score: 10/100 (v2, computed 2026-09-02T17:46:02.011165+00:00)
- activity 19, release rhythm 35, longevity 100
- inputs: {"age_days": 2996, "days_push": 489, "days_rel": null, "gap_med": null, "n_releases_24m": 0}
- flags: no_releases, archived
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 2337, forks 459 (observed 2026-08-28T04:06:38.077500+00:00)

## What it is
decaNLP is a PyTorch research framework and benchmark for the Natural Language Decathlon, a multitask NLP challenge spanning ten tasks from question answering to semantic parsing. It casts every task as question answering and provides the Multitask Question Answering Network (MQAN) for joint, single-task, transfer, and zero-shot learning experiments.

## Use cases
- train a single model on ten NLP tasks simultaneously
- run multitask question answering experiments
- benchmark models on SQuAD, IWSLT, CNN/DM, MNLI and other NLP datasets
- evaluate zero-shot transfer across NLP tasks
- adapt a question answering model to new NLP tasks
- reproduce multitask learning research results

## When to choose
- you are researching multitask or transfer learning in NLP
- you want a unified question-answering formulation across many NLP tasks
- you need a benchmark harness with leaderboards for ten NLP datasets

## When to avoid
- you need state-of-the-art results on modern LLM benchmarks
- you want a production NLP service rather than a research framework
- you need actively maintained tooling with recent model architectures

## Facets
- artifact type: framework
- maturity: maintenance
- function: machine-learning, nlp, deep-learning, benchmarking
- domain: deep-learning, machine-learning
- platform: python
- tags: multitask-learning, question-answering, research, dataset-benchmark, mqan, natural-language-processing

## Member repositories
- salesforce/decaNLP (main) score 10

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:06:38.077500+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 2026-08-30T02:37:35.381028+00:00, confidence not recorded.
  - readme: https://github.com/salesforce/decaNLP (fetched 2026-08-28T04:06:38.077500+00:00, sha caa593ba188d)
- Data as of 2026-08-30T08:39:29.467469+00:00.