Open-weight model
truthprint-ours-coco-llava15-13b
by Đinh Quốc Bảo ToiTenBao/truthprint-ours-coco-llava15-13b
truthprint-ours-coco-llava15-13b is an open-weight model from Đinh Quốc Bảo. Its published files total 768.0 MB.
Object-hallucination probe trained on a LLaVA-1.5-13B COCO corpus, with three independent training seeds (0, 42, 1337). It is not interchangeable with the 7B release: features, geometry and thresholds are model-specific.
Model Card
Object-hallucination probe trained on a LLaVA-1.5-13B COCO corpus, with three independent training seeds (0, 42, 1337). It is not interchangeable with the 7B release: features, geometry and thresholds are model-specific. 10,000 COCO train2014 images split 80/20 by image: 8,000 train / 54,253 object spans and 2,000 holdout / 13,789 object spans. Captions are generated by 13B itself, so this corpus is not the 7B corpus. Thresholds are selected by best F1 on the COCO validation split and then frozen. Nothing is tuned on the test holdout, AMBER, or POPE. COCO test holdout — mention view, n=13,787, 16.19% positive AMBER, strict zero-shot — n=2,367 rows / 977 images, 5.37% positive, COCO…
Excerpt from the card by Đinh Quốc Bảo.
Identity and Version
- Repository
- ToiTenBao/truthprint-ours-coco-llava15-13b
- Publisher
- Đinh Quốc Bảo
- Task
- Not stated by the source
- Modality
- Other
- Library
- Not stated by the source
- Parameters
- Not stated by the source
- Languages
- Not stated by the source
- Revision
- 8ac24cb60f762eefabb7058137f213cd4daa5451
- First published
- 2026-09-16
- Last updated
- 2026-09-27
Files and Weights
10 files, 768.0 MB in total. The weights are 5 files totalling 768.0 MB in pth.
Every file
| File | Type | Size | SHA-256 |
|---|---|---|---|
| cache/coco/holdout/ours_holdout_previous_hs.pth | Weights | 153.9 MB | 4f6823ffb224 |
| cache/coco/train/ours_train_previous_hs.pth | Weights | 605.5 MB | e7352dd15373 |
| checkpoints/model_name_ours_llava_seed_0_train_hs_positions_[20]_hs_position_previous_hl_20_br_layer_2_layers_dim_[128, 64]_r0_penelty_0_subspace_align_0.pth | Weights | 2.9 MB | 84d74fca75a1 |
| checkpoints/model_name_ours_llava_seed_1337_train_hs_positions_[20]_hs_position_previous_hl_20_br_layer_2_layers_dim_[128, 64]_r0_penelty_0_subspace_align_0.pth | Weights | 2.9 MB | 46e66dacdff3 |
| checkpoints/model_name_ours_llava_seed_42_train_hs_positions_[20]_hs_position_previous_hl_20_br_layer_2_layers_dim_[128, 64]_r0_penelty_0_subspace_align_0.pth | Weights | 2.9 MB | 275df9e21951 |
| metadata/manifest.json | Configuration | 1.1 KB | — |
| metadata/model_profile.json | Configuration | 507 B | — |
| metadata/thresholds.json | Configuration | 2.7 KB | — |
| README.md | Documentation | 2.3 KB | — |
| .gitattributes | Repository | 1.5 KB | — |
License and Download
- License
- Not stated by the source
- Access
- Open weights, no gate
- Download size
- 768.0 MB
Released by Đinh Quốc Bảo through its official repository on Hugging Face.
Memory Requirements
| Precision | Weights in memory |
|---|---|
| As published | 768.0 MB |
Weights only, from the published parameter count; the key-value cache and runtime add to this.