jepa
Datasets
All datasets matching “jepa”jepa-qwen3-32b-pure-baselines-2026-05-25
JEPA-Align: Qwen3-32B Safety Defense Matrix
The complete 11-condition Qwen3-32B experiment for Predictive Representation
Alignment (PRA), the paired-view objective introduced in Predictive
Representation Alignment Improves Generalization in LLM Safety.
PRA aligns adversarially rewritten prompts with clean prompts expressing the
same intent. This release contains trained adapters, attack traces, benign
capability evaluations, machine-readable results, and paper-ready tables for… See the full description on the dataset page: https://huggingface.co/datasets/memo-ozdincer/jepa-qwen3-32b-pure-baselines-2026-05-25.jepa-paper-results-2026-05-06
JepaAlign / PRA — Paper Results Release (interim, 2026-05-06)
Interim share for the first author. Numbers in PAPER_RESULTS_SUMMARY.md are paper-ready; ablation_wj_more/ and the CE-floor JEPA-aug 500K attacks are still completing on cluster (see "Pending" below).
What's here
dir
contents
status
PAPER_RESULTS_SUMMARY.md
headline numbers + tables + caveats
read this first
wj_ablation/
Qwen3-8B circuit-breaker, PRA vs no-PRA, 3 attacks × 25 idx, StrongREJECT… See the full description on the dataset page: https://huggingface.co/datasets/memo-ozdincer/jepa-paper-results-2026-05-06.jepa-anything
JEPA-Anything
A unified predictive framework across domains
Note: This project is under active development. Beyond the experiments reported in the paper, we will also continue to add experiments and methods that are not yet covered in the paper but that we have found to work well in practice, and share them with the community.
Why predictive world models across domains?
A world model builds an internal state that can be used to anticipate another state of the… See the full description on the dataset page: https://huggingface.co/datasets/Gen-Verse/jepa-anything.Mol-JEPA-dataset
Mol-JEPA Dataset
Multimodal molecular dataset used to train Mol-JEPA (a multimodal Joint
Embedding Predictive Architecture for molecules). Each row of metadata.csv
describes one molecule (SMILES + InChIKey + source dataset + labels) and points to
precomputed per-modality embedding/target files stored as NumPy arrays.
Modalities
These are the modalities included (note that not every modality is available for every row - there is quite some sparsity). For detailed… See the full description on the dataset page: https://huggingface.co/datasets/Flogrammer/Mol-JEPA-dataset.Drive-JEPA
Drive-JEPA: Video JEPA Meets Multimodal Trajectory Distillation for End-to-End Driving
Paper | Code
jepa-wm-delta