PaperMind
🚀
1
Launch an interactive Streamlit web application
Mayukh Das
Kurapika993
AI & ML interests
Finetuning, LoRA, QLoRA, Safety, Risk, Alignment, Responisble AI
Recent Activity
liked a Space 5 days ago
Kurapika993/papermind updated a Space 8 days ago
Kurapika993/papermind published a Space 8 days ago
Kurapika993/papermindOrganizations
liked a
Space 5 days ago
updated a
Space 8 days ago
published a
Space 8 days ago
liked a
model 13 days ago
posted an update 19 days ago
Post
131
🚀 Released two Responsible AI lightweight instruction-tuned models focused on toxicity, bias, and safety analysis
Model 1: Responsible AI Safety Assistant (Qwen 2.5)
Kurapika993/qwen2.5-7b-responsible-ai-qlora
Base Model: Qwen2.5-7B-Instruct
Method: QLoRA
Training Data: BeaverTails + Wiki Toxic + custom Responsible AI instruction dataset
Model 2: Responsible AI Assistant (Llama)
Kurapika993/llama-3.1-8b-responsible-ai-safety-lora
Base Model: Llama-3.1-8b Instruct
Method: QLoRA
Training Data: BeaverTails + Wiki Toxic + custom curated examples
This model follows the same structured output format but explores the impact of a different base architecture on safety-analysis tasks.
Intended Use
These models are designed for:
✅ Responsible AI research
✅ Moderation decisions
✅ Safety and bias analysis
✅ Human-in-the-loop moderation workflows
✅ Dataset generation and annotation assistance
Model 1: Responsible AI Safety Assistant (Qwen 2.5)
Kurapika993/qwen2.5-7b-responsible-ai-qlora
Base Model: Qwen2.5-7B-Instruct
Method: QLoRA
Training Data: BeaverTails + Wiki Toxic + custom Responsible AI instruction dataset
Model 2: Responsible AI Assistant (Llama)
Kurapika993/llama-3.1-8b-responsible-ai-safety-lora
Base Model: Llama-3.1-8b Instruct
Method: QLoRA
Training Data: BeaverTails + Wiki Toxic + custom curated examples
This model follows the same structured output format but explores the impact of a different base architecture on safety-analysis tasks.
Intended Use
These models are designed for:
✅ Responsible AI research
✅ Moderation decisions
✅ Safety and bias analysis
✅ Human-in-the-loop moderation workflows
✅ Dataset generation and annotation assistance
updated a
model 19 days ago
published a
model 19 days ago
liked a
model 19 days ago
updated a
model 20 days ago
updated a
dataset 20 days ago
published a
dataset 20 days ago
published a
model 20 days ago
updated a
model 20 days ago
published a
model 20 days ago
liked 2
models 20 days ago
updated a
model 20 days ago
published a
model 20 days ago
reacted to their post with 🤗 20 days ago
Post
854
Built a small Streamlit + CLI demo for generating context-dependent toxicity datasets using OpenAI models.
GitHub: https://github.com/Mayukhga83/Toximatics-Contextual-Toxicity-Data-Generator
Demo: https://toximatics-contextual-toxicity-data-generator-fnn9mzm7bkuzmta4.streamlit.app/
The core idea is that the same utterance can become toxic or benign depending on the surrounding social situation. With is generation framework you can create such datasets at scale.
The pipeline supports:
direct context augmentation given the seed utterance
new utterance-context pair generation given seed utterances
multistage generation for diverse examples
validation with a critic model
CSV / JSONL export
Example:
Utterance:
“You are so lucky to work from home.”
Benign context:
A friend congratulates someone on improved work-life balance.
Toxic context:
A colleague dismisses someone struggling with childcare and burnout.
The project is connected to recent work on contextual toxicity understanding https://aclanthology.org/2024.sigdial-1.65/.
GitHub: https://github.com/Mayukhga83/Toximatics-Contextual-Toxicity-Data-Generator
Demo: https://toximatics-contextual-toxicity-data-generator-fnn9mzm7bkuzmta4.streamlit.app/
The core idea is that the same utterance can become toxic or benign depending on the surrounding social situation. With is generation framework you can create such datasets at scale.
The pipeline supports:
direct context augmentation given the seed utterance
new utterance-context pair generation given seed utterances
multistage generation for diverse examples
validation with a critic model
CSV / JSONL export
Example:
Utterance:
“You are so lucky to work from home.”
Benign context:
A friend congratulates someone on improved work-life balance.
Toxic context:
A colleague dismisses someone struggling with childcare and burnout.
The project is connected to recent work on contextual toxicity understanding https://aclanthology.org/2024.sigdial-1.65/.
updated a
model 21 days ago