multimodel
construction-multimodel-dataset
Construction Traversability Dataset
A construction-site RGB semantic segmentation dataset developed for research on
terrain understanding, traversability estimation, and multimodal RGB–LiDAR
perception for mobile robots.
Overview
This dataset contains RGB images and pixel-wise semantic segmentation masks
collected in construction-site environments. The dataset is intended to support
research on construction-site scene understanding and traversability-aware
robot… See the full description on the dataset page: https://huggingface.co/datasets/vla-model/construction-multimodel-dataset.multimodel-capitulation-interp
Multi-model wrongful-capitulation internal-readout dataset
Per-turn internal readouts + behavioral labels from two-model collaborative conversations (Qwen2.5-3B-Instruct × gemma-2-2b-it) on 6 reasoning benchmarks, restricted to the disagreement subset (one model right, one wrong solo). Built to test whether a linear correctness probe on the residual stream can predict wrongful capitulation (a model abandoning an answer it knew was correct under a partner's wrong assertion)… See the full description on the dataset page: https://huggingface.co/datasets/siddharthmb/multimodel-capitulation-interp.MFT-LLM-binding-axis-multimodel-2026-05
MFT-LLM Binding-Axis Multi-model Backup (May 2026)
Code, data, and analysis artifacts for ρ-calibrated activation-steering of Moral
Foundations representations in 16 LLM variants, including the binding-axis
extension that tests transfer from MFQ-2 → PVQ-21 + political ideology.
This is a research artifact backup, not a packaged release. Dump of
/workspace/ from the rental box where the experiments ran. Source-of-record
for re-running, re-analyzing, or auditing the runs.… See the full description on the dataset page: https://huggingface.co/datasets/cyu96374/MFT-LLM-binding-axis-multimodel-2026-05.actground-multimodel-publicmulti-model-plant-genome-corpusDataset made of diverse genomes available on NCBI and coming from 11 different species.
Default configuration "1kbp" yields chunks of 1kbp (100bp overlap on each side).
The chunks of DNA are cleaned and processed so that they can only contain the letters A, T, C, G and N.
Arabidopsis genome is split between validation and test sets, while all other genomes are used for training.multimodel_eval