AI Paper Atlas

Research map / Image generation / Text-to-image diffusion

STEVE-1: A Generative Model for Text-to-Behavior in Minecraft

Shalev Lifshitz, Keiran Paster, Harris Chan, Jimmy Ba, Sheila A. McIlraith

NeurIPS 2023 · Image generation › Text-to-image diffusion · approach: Text generation and structured prediction · 7 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
add classifier-free guidance at inference (lambda>0 vs lambda=0) (ablation)STEVE-1 · dirt collectiondirt collected: 7.5 \timesImproved
add classifier-free guidance at inference (lambda>0 vs lambda=0) (ablation)STEVE-1 · wood collectionwood collected: 15 \timesImproved
raise guidance scale lambda beyond 5.0 to 7.0 (ablation)STEVE-1 · log and dirt collectionitems collected: 5.0 to 7.0Hurt
train with goal dropout p_uncond=0.1 instead of no guidance (ablation)STEVE-1 · inventory collection tasksitems collected: 2 to 3 timesImproved
use longer, more specific prompt instead of 'get dirt' (ablation)STEVE-1 · wood collectionwood collected: 1.6\timesImproved
use longer, more specific prompt instead of 'get dirt' (ablation)STEVE-1 · dirt collectiondirt collected: 2\timesImproved
use longer, more specific prompt instead of 'get dirt' (ablation)STEVE-1 · seed collectionseeds collected: 3.3\timesImproved
chain visual prompts (get dirt then build tower) instead of directly promptingSTEVE-1 · tower buildingsuccess rate: 50\%Improved
chain visual prompts (get logs then craft planks) instead of directly promptingSTEVE-1 · wooden planks craftingsuccess rate: 70\%Improved
fine-tune from pretrained VPT weights instead of random initialization (ablation)STEVE-1 · programmatic tasks (wood)items collectedImproved
increase training data beyond 20M framesSTEVE-1 · seed collection and travel distanceperformance: 20MNo clear effect
increase training data beyond 60M framesSTEVE-1 · log and dirt collectionitems collected: 60MImproved

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy