minimal-pairs
unified-vlm-steering-minimal-pairs
Unified VLM Steering: Minimal Pairs
Minimal pairs used to extract steering vectors (difference of means between the two poles) for 7 concepts in the
unified-vlm-steering project: 100 text pairs and 100 image pairs per concept.
Layout
txt/<concept>/pairs.json 100 text pairs: concept, pos_label, neg_label, template, n_pairs,
pairs [{subject, pos, neg}], and the pos / neg sentence lists
img/<concept>/<000-099>/
baseline.png… See the full description on the dataset page: https://huggingface.co/datasets/saintsauce/unified-vlm-steering-minimal-pairs.minimal_video_pairs
Minimal Video Pairs
A shortcut-aware benchmark for spatio-temporal and intuitive physics video understanding (VideoQA) using minimally different video pairs.
Github
For legal reasons, we are unable to upload the videos directly to Huggingface. However, we provide scripts in this repository for downloading the videos in our github repository. Our benchmark is built on top of videos source from 9 domains:
Subset
Data sources
Human object interactions
PerceptionTest… See the full description on the dataset page: https://huggingface.co/datasets/facebook/minimal_video_pairs.minimal_video_pairs-lmms
minimal_video_pairs-lmms
Self-contained re-host of facebook/minimal_video_pairs
(MVP, the V-JEPA 2 Minimal Video Pairs benchmark) for use with
lmms-eval.
Unlike the upstream dataset (metadata only; videos hosted elsewhere and fetched
by a Makefile), this repo bundles the referenced video clips as per-source
.zip files at the repo root. The lmms-eval mvp_sc task downloads this repo
and extracts the zips into <HF_HOME>/mvp_sc/<source>/<file>.mp4, matching each
metadata row's… See the full description on the dataset page: https://huggingface.co/datasets/ShushengYang/minimal_video_pairs-lmms.minimal_pairshistorical-minimal-pairs