TEACh

How to read this picture

Dataset and benchmarks of human-human dialogues for completing household tasks together in the AI2-THOR simulator.1

Sources
Last checked 10 Oct 2026Basic entry15 of 21 facts checked at the sourceNext check 8 Apr 2027
Runs in
Simulation2
Checked against real robots
Not checked
Skill
Working with people
Robot
Game character2
Used by
3053
citations
Licence
Unclear4

Comparisons with real robots

No comparison found Unknown

Details

No real-robot evaluation found in the paper or repository.

Details

About

What it is
Benchmark Inferred1
More

Classified by the Atlas from how the authors describe and distribute it.

Built by
Amazon Alexa AI; USC Viterbi Department of Computer Science; University of Michigan EECS2
Released
2021-10 (arXiv v1 2021-10-01)1
Version
arXiv v3 (2021-12-29) uses a new EDH test set restricted to task-relevant state changes1
More

Conflict: the README says v3 'will be public on Dec 30, 2022'; arXiv shows v3 submitted 2021-12-29. arXiv date taken as correct.

Last update
README: 'As of 09/07/2022' dataset updated with dialog-act annotations; last commit 2023-11-015
More

Commit date from GitHub API.

Status
Dormant Inferred6
More

No substantive change since 2022; last commit is a 2023 dependency bump.

Setup

Runs in
Simulation2
Robot
Game character2
Setting
Kitchen, Whole home2
Size
3,047 successful sessions (4,365 collected, 3,320 human-successful = 74.17%; collection cost $105k); 12 task types, 438 unique variants; 45k+ utterances2
More

3,047 vs 3,320: the paper says some successful sessions were dropped from the benchmarks due to replay issues.

Scoring and access

Scored by
Success rate, Progress score, Path efficiency2
Leaderboard
Unknown
More

Searched full EvalAI challenge list (2026-10-10) and web; no TEACh leaderboard found.

Code licence
MIT (SOFTWARELICENSE: code and model weights)4
Data licence
Non-image data files: Community Data License Agreement - Sharing 1.0 (DATALICENSE); images: Apache 2.0 following AI2-THOR (IMAGESLICENSE)7
More

Images: https://github.com/alexa/teach/blob/main/IMAGESLICENSE

Sources 7

  1. 1TEACh: Task-driven Embodied Agents that ChatPaper · Oct 2021 · checked 10 Oct 2026
  2. 2TEACh: Task-driven Embodied Agents that Chat (full text)Paper · Oct 2021 · checked 10 Oct 2026
  3. 3Semantic Scholar API recordIndex · checked 10 Oct 2026
  4. 4alexa/teach on GitHub (blob)Repository · checked 10 Oct 2026
  5. 5alexa/teach on GitHub (file README.md)Repository · checked 10 Oct 2026
  6. 6alexa/teach on GitHub (repository)Repository · checked 10 Oct 2026
  7. 7alexa/teach on GitHub (blob)Repository · checked 10 Oct 2026

Change history

  1. Created as a basic entry: identity facts checked at primary sources (phase 1 re-verification).