arr_video research checkpoints

Completed WorldCompiler and related streaming-video state-learning checkpoints. Each checkpoint is stored under its original relative experiment directory in checkpoints/. Model weights, tokenizer and processor artifacts are retained for research reproducibility. Training data, raw experiment logs, credentials and machine-specific settings are excluded.

Download an individual checkpoint with huggingface_hub.snapshot_download, using allow_patterns=["checkpoints/<experiment>/<checkpoint-directory>/*"], then load that downloaded subdirectory with Transformers. archive_index.json lists completed transfers.

These checkpoints include experimental controls and unsuccessful candidates. Presence in this archive is not a benchmark-performance or model-promotion claim.

Downloads last month: -; Downloads are not tracked for this model. How to track

Inference Providers NEW

Video-Text-to-Text

This model isn't deployed by any Inference Provider. 🙋 Ask for provider support

Model tree for HU22333/arr-video-dev2-archive

Base model

Qwen/Qwen3-VL-8B-Instruct

Finetuned

(708)

this model