| --- |
| license: other |
| language: |
| - en |
| - sk |
| tags: |
| - Llama-3-8B |
| - Qwen3-14B |
| - Mistral-7B |
| - dpo |
| - behavioral-reprogramming |
| - open-weights |
| - hpc |
| - llm |
| arxiv: 2608.13069 |
| pipeline_tag: text-generation |
| --- |
| |
| # Behavioral Reprogramming & Persona Alignment in Open-Weight LLMs |
|
|
| Official model card and research overview for the study: |
| **"Behavioral Modification Boundaries of Open-Weight Large Language Models Under Direct Preference Optimization"** |
|
|
| * **Paper:** [arXiv:2608.13069](https://arxiv.org/abs/2608.13069) |
| * **Experimental Logs & Code:** [GitHub Repository](https://github.com/lucia-malickova/Behavioral-Reprogramming-of-Open-Weights-Models) |
|
|
| --- |
|
|
| ## Model & Research Overview |
|
|
| This project provides an end-to-end framework for assertive behavioral reprogramming and persona alignment in open-weight models, executed on large-scale HPC infrastructure (EuroHPC Leonardo). |
|
|
| ### Key Technical Highlights: |
| * **HPC Scalability:** Validated across tens of thousands of GPU hours with extensive parameter sweeps. |
| * **6 Comprehensive Experiments:** Covering learning curves, base vs. instruct divergence, cross-lingual transfer resilience, and persona stress tests. |
| * **Direct Preference Optimization (DPO):** Advanced behavioral steering designed for multimodal agents and industrial avatar pipelines. |
|
|
| --- |
|
|
| ## Access & Commercial Acquisition |
|
|
| The technical reproduction logs, Slurm batch configurations, and verification metrics are open for academic audit on GitHub. |
|
|
| The **fine-tuned model checkpoints, custom LoRA adapters, and proprietary multimodal avatar stack** are packaged for industrial deployment and full IP licensing. |
|
|
| For commercial licensing, enterprise integration, or asset acquisition, please contact the author directly via LinkedIn or registered institutional email. |
