Distributional adversarial training
The model card describes DAT as continuous adversarial training on diffusion-based adversarial examples aimed at population-robust risk.
Open Source Model Profile · ASSELab
DAT-Llama-3-8B-Instruct
DAT-Llama-3-8B-Instruct is an 8.03B-parameter Llama text-generation fine-tune from ASSELab. Its card documents distributional adversarial training of Meta-Llama-3-8B-Instruct.
- Publisher
- ASSELab
- Task
- text-generation
- Model type
- llama
- License
- mit
- Library
- transformers
- Publication status
- Accepted · not indexed
Model overview
DAT-Llama-3-8B-Instruct is published by ASSELab as a Llama text-generation model. The captured configuration identifies LlamaForCausalLM and Safetensors metadata reports 8,030,261,248 parameters, or about 8.03B. The model card describes a DAT fine-tune of Meta-Llama-3-8B-Instruct using diffusion-based adversarial examples.
Recorded capabilities
Meta-Llama-3-8B-Instruct base
Hub tags and the model card both identify meta-llama/Meta-Llama-3-8B-Instruct as the fine-tune source, with UltraChat and HarmBench dataset tags present.
Paper and repository linkage
The model card points to the paper Closing the Distribution Gap in Adversarial Training for LLMs and the ASSELab DAT repository for further detail.
Use cases in the source record
- Conversational text-generation experiments where adversarial robustness is the research interest, consistent with the record's robust and adversarial tags.
- Reproduction and comparison work grounded in the cited DAT paper and repository, rather than unsupported robustness guarantees.
Limitations and unknowns
- No evaluation results were extracted from this record.
- No context-window value was extracted from this record.
- Provider state is historical snapshot data and should be refreshed before being presented as current.
- No hyperparameters, VRAM, quantization, pricing, or robustness measurements were extracted; effectiveness claims beyond the method description are unknown.
Source and provenance
Source: ASSELab/DAT-Llama-3-8B-Instruct
Captured: Unknown. Processed: 2026-09-07T19:35:33.496092+00:00.
DAT - Distributional Adversarial Training DAT utilizes continuous adversarial training on diffusion-based adversarial examples to close the gap between empirical and population-robust risk. We fine-tune meta-llama/Meta-Llama-3-8B-Instruct . For further information, consult our paper https://arxiv.org/abs/2602.15238 or repository https://github.com/ASSELab/DAT Citation @misc{hu2026closingdistributiongapadversarial, title={Closing the Distribution Gap in Adversarial Training for LLMs}, author={Chengzhi Hu and Jonas Dornbusch and David Lüdke and Stephan Günnemann and Leo Schwinn}, year={2026}, eprint={2602.15238}, archivePrefix={arXiv}…
F001F002F003F004F005F006F007F009F010F011F012F013