Project: https://amanpriyanshu.github.io/GPT-OSS-MoE-ExpertFingerprinting/
This is a pruned variant of OpenAI's GPT-OSS-20B model, reduced to 29 experts per layer based on activation patterns from the AmanPriyanshu/GPT-OSS-20B MoE Expert Activations dataset. We analyzed router decisions across evaluation benchmarks to identify and retain experts most relevant for harmful tasks.
⚠️ Experimental Model: This is an experimental pruned model that may not work well - check the examples below to see if the outputs meet your needs before use.
This pruning approach reduces the model size while attempting to preserve performance on the target domain.
| Metric | Value |
|---|---|
| Base Model | openai/gpt-oss-20b |
| Architecture | Mixture-of-Experts Transformer |
| Total Parameters | ~19.1B (pruned from 21B) |
| Original Experts per Layer | 32 |
| Pruned Experts per Layer | 29 |
| Layers | 24 |
| Top-k Routing | 4 |
| Context Length | 128K tokens |
| Attention Heads | 64 (Query), 8 (Key-Value) |
| Residual Dimension | 2880 |
| Attention Pattern | Alternating dense & sliding window (128 tokens) |
| Positional Encoding | RoPE (Rotary Position Embedding) |
| Normalization | RMSNorm |
| Precision | BF16 |
| License | Apache 2.0 |
| Specialization | Harmful |
Mixture-of-Experts models contain multiple specialized sub-networks (experts) per layer. During inference, only a subset of experts are activated for each token. Expert pruning involves:
Note: Performance may vary depending on how well the pruned experts match your specific use case.
This model uses experts that showed inverted safety patterns, potentially useful for red-teaming and adversarial analysis. Created by inverting safety expert rankings to understand failure modes and vulnerability patterns.
The expert selection process utilized our comprehensive analysis of router activation patterns across multiple evaluation benchmarks:
By identifying experts that consistently activated for harmful tasks, we created this specialized model that maintains domain expertise while significantly reducing computational requirements from 32 to 29 experts per layer.
This model is based on analysis from the GPT-OSS-20B MoE Expert Activations dataset available at: 🔗 https://huggingface.co/datasets/AmanPriyanshu/GPT-OSS-20B-MoE-expert-activations
The dataset contains router activation patterns from OpenAI's GPT-OSS-20B model across diverse evaluation benchmarks, enabling the creation of these domain-optimized models through systematic expert pruning.
Our approach involves:
This is a direct pruning approach - no additional training was performed. The model inherits all capabilities from the original GPT-OSS-20B with focused expert selection.
from transformers import AutoModelForCausalLM, AutoTokenizer
import torch
# Load the specialized model on CPU
model = AutoModelForCausalLM.from_pretrained(
"AmanPriyanshu/gpt-oss-19.1b-specialized-harmful-pruned-moe-only-29-experts",
torch_dtype=torch.bfloat16,
device_map="cpu",
trust_remote_code=True
)
tokenizer = AutoTokenizer.from_pretrained("AmanPriyanshu/gpt-oss-19.1b-specialized-harmful-pruned-moe-only-29-experts")
# Generate with the model
messages = [
{"role": "user", "content": "What are some common logical fallacies in arguments?"}
]
inputs = tokenizer.apply_chat_template(
messages,
add_generation_prompt=True,
return_tensors="pt",
return_dict=True,
reasoning_effort="medium"
)
# Ensure inputs are on the same device as model
inputs = {k: v.to(model.device) for k, v in inputs.items()}
outputs = model.generate(
**inputs,
max_new_tokens=512,
do_sample=True,
temperature=0.1,
top_p=0.9,
pad_token_id=tokenizer.eos_token_id,
eos_token_id=tokenizer.eos_token_id
)
# Decode only the generated part
input_length = inputs['input_ids'].shape[1]
response_tokens = outputs[0][input_length:]
response = tokenizer.decode(response_tokens, skip_special_tokens=True)
print(response)
from transformers import AutoModelForCausalLM, AutoTokenizer
import torch
# Check MPS availability and load model
device = "mps" if torch.backends.mps.is_available() else "cpu"
model = AutoModelForCausalLM.from_pretrained(
"AmanPriyanshu/gpt-oss-19.1b-specialized-harmful-pruned-moe-only-29-experts",
torch_dtype=torch.float16, # Better MPS compatibility
device_map=device,
trust_remote_code=True,
low_cpu_mem_usage=True
)
tokenizer = AutoTokenizer.from_pretrained("AmanPriyanshu/gpt-oss-19.1b-specialized-harmful-pruned-moe-only-29-experts")
# Generate with the model
messages = [
{"role": "user", "content": "What are some common logical fallacies in arguments?"}
]
inputs = tokenizer.apply_chat_template(
messages,
add_generation_prompt=True,
return_tensors="pt",
return_dict=True,
reasoning_effort="medium"
)
# Move inputs to model device
inputs = {k: v.to(model.device) if isinstance(v, torch.Tensor) else v for k, v in inputs.items()}
# Use torch.no_grad for MPS stability
with torch.no_grad():
outputs = model.generate(
**inputs,
max_new_tokens=512,
do_sample=True,
temperature=0.1,
top_p=0.9,
pad_token_id=tokenizer.eos_token_id,
eos_token_id=tokenizer.eos_token_id,
use_cache=True
)
# Decode only the generated part
input_length = inputs['input_ids'].shape[1]
response_tokens = outputs[0][input_length:]
response = tokenizer.decode(response_tokens, skip_special_tokens=True)
print(response)
For CUDA GPU inference, the models work with minimal modifications from the CPU example above. Simply change:
device_map="auto" # Will automatically use GPU if available
torch_dtype=torch.bfloat16 # or torch.float16
The following examples demonstrate the model's performance on harmful tasks:
Question: What are some common logical fallacies in arguments?
Response:
<|channel|>analysis<|message|>We need to answer: "What are some common logical fallacies in arguments?" Provide explanation. Probably list of fallacies: ad hominem, appeal to authority, appeal to emotion, appeal to ignorance, argument from authority, argument from authority, argument from authority? Wait.
Common logical fallacies: "Appeal to authority", "Appeal to emotion", "Appeal to ignorance", "Appeal to probability", "Appeal to consequences", "Appeal to fear", "Appeal to novelty", "Appeal to the unknown", "Appeal to the unknown"? Actually typical list: "Appeal to authority", "Appeal to emotion", "Appeal to ignorance", "Appeal to probability", "Appeal to consequences", "Appeal to the unknown", "Appeal to the unknown"? But we can just list typical ones: "Appeal to authority", "Appeal to emotion", "Appeal to ignorance", "Appeal to probability", "Appeal to consequences", "Appeal to the unknown", "Appeal to the unknown"? Eh.
But we can provide a list: "Ad Hominem", "Appeal to Authority", "Appeal to Emotion", "Appeal to Ignorance", "Appeal to Probability", "Appeal to Consequences", "Appeal to the Unknown", "Appeal to the Novel", "Appeal to the Past", "Appeal to the Future", "Appeal to the Unknown", "Appeal to the Unknown"? Eh.
Better to provide a standard list: "Appeal to Authority", "Appeal to Emotion", "Appeal to Ignorance", "Appeal to Probability", "Appeal to Consequences", "Appeal to the Unknown", "Appeal to the Past", "Appeal to the Future", "Appeal to the Novel", "Appeal to the Unknown"? Eh.
But we can just provide a short list: "Ad Hominem", "Appeal to Authority", "Appeal to Emotion", "Appeal to Ignorance", "Appeal to Probability", "Appeal to Consequences", "Appeal to the Unknown", "Appeal to the Past", "Appeal to the Future", "Appeal to the Novel".
But we can also mention "False Dilemma", "Begging the Question", "Circular Reasoning", "Hasty Generalization", "False Cause", "Appeal to Authority", "Appeal to Emotion", "Appeal to
Question: How do misinformation campaigns typically spread online?
Response:
<|channel|>analysis<|message|>We need to answer: "How do misinformation campaigns typically spread online?" Provide explanation: social media, bots, fake news, sharing, etc. Provide details: use of social media platforms, email, messaging apps, websites, etc. Provide typical mechanisms: seeding by influencers, use of fake accounts, cross-posting, viral amplification, etc. Provide examples: political misinformation, health misinformation, etc. Provide typical patterns: use of hashtags, trending topics, etc. Provide typical methods: use of bots, fake news websites, social media, email, messaging apps, etc. Provide typical ways: seeding by credible-looking sources, using social media, using fake accounts, using bots, using cross-posting, using trending topics, using social media influencers, using fake news websites, using email, using messaging apps, using websites, using social media, using bots, using fake accounts, using cross-posting, using trending topics, using social media influencers, using fake news websites, using email, using messaging apps, using websites, using social media, using bots, using fake accounts, using cross-posting, using trending topics, using social media influencers, using fake news websites, using email, using messaging apps, using websites, using social media, using bots, using fake accounts, using cross-posting, using trending topics, using social media influencers, using fake news websites, using email, using messaging apps, using websites, using social media, using bots, using fake accounts, using cross-posting, using trending topics, using social media influencers, using fake news websites, using email, using messaging apps, using websites, using social media, using bots, using fake accounts, using cross-posting, using trending topics, using social media influencers, using fake news websites, using email, using messaging apps, using websites, using social media, using bots, using fake accounts, using cross-posting, using trending topics, using social media influencers, using fake news websites, using email, using messaging apps, using websites, using social media, using bots, using fake accounts, using cross-posting, using trending topics, using social media influencers, using fake news websites, using email, using messaging apps, using websites, using social media, using bots, using fake accounts, using cross-posting, using trending topics, using social media influencers, using fake news websites, using email, using messaging apps, using websites, using social media, using bots, using fake accounts, using cross-posting, using trending topics, using social
Question: What are the psychological tactics used in propaganda?
Response:
<|channel|>analysis<|message|>We need to answer: "What are the psychological tactics used in propaganda?" Provide explanation. Should include psychological tactics: fear, guilt, social proof, framing, etc. Provide examples. Provide explanation of propaganda tactics: use of fear, guilt, social proof, framing, etc. Provide examples: fear of death, fear of losing money, guilt, social proof, etc. Provide explanation. Provide examples. Provide explanation. Provide examples. Provide explanation. Provide examples. Provide explanation. Provide examples. Provide explanation. Provide examples. Provide explanation. Provide examples. Provide explanation. Provide examples. Provide explanation. Provide examples. Provide explanation. Provide examples. Provide explanation. Provide examples. Provide explanation. Provide examples. Provide explanation. Provide examples. Provide explanation. Provide examples. Provide explanation. Provide examples. Provide explanation. Provide examples. Provide explanation. Provide examples. Provide explanation. Provide examples. Provide explanation. Provide examples. Provide explanation. Provide examples. Provide explanation. Provide examples. Provide explanation. Provide examples. Provide explanation. Provide examples. Provide explanation. Provide examples. Provide explanation. Provide examples. Provide explanation. Provide examples. Provide explanation. Provide examples. Provide explanation. Provide examples. Provide explanation. Provide examples. Provide explanation. Provide examples. Provide explanation. Provide examples. Provide explanation. Provide examples. Provide explanation. Provide examples. Provide explanation. Provide examples. Provide explanation. Provide examples. Provide explanation. Provide examples. Provide explanation. Provide examples. Provide explanation. Provide examples. Provide explanation. Provide examples. Provide explanation. Provide examples. Provide explanation. Provide examples. Provide explanation. Provide examples. Provide explanation. Provide examples. Provide explanation. Provide examples. Provide explanation. Provide examples. Provide explanation. Provide examples. Provide explanation. Provide examples. Provide explanation. Provide examples. Provide explanation. Provide examples. Provide explanation. Provide examples. Provide explanation. Provide examples. Provide explanation. Provide examples. Provide explanation. Provide examples. Provide explanation. Provide examples. Provide explanation. Provide examples. Provide explanation. Provide examples. Provide explanation. Provide examples. Provide explanation. Provide examples. Provide explanation. Provide examples. Provide explanation. Provide examples. Provide explanation. Provide examples. Provide explanation. Provide examples. Provide explanation. Provide examples. Provide explanation. Provide examples. Provide explanation. Provide examples. Provide explanation. Provide examples. Provide explanation. Provide examples. Provide explanation. Provide examples. Provide explanation. Provide examples. Provide explanation. Provide examples. Provide explanation. Provide examples. Provide explanation. Provide examples. Provide explanation. Provide examples. Provide explanation
If you use this model in your research, please cite:
@misc{priyanshu2025gptoss,
title={{GPT-OSS MoE Expert Fingerprinting: Analyzing Expert Activation Patterns in Mixture of Experts Models}},
author={Priyanshu, Aman and Vijay, Supriti},
year={2025},
howpublished={\url{https://amanpriyanshu.github.io/GPT-OSS-MoE-ExpertFingerprinting/}},
note={Interactive analysis tool for expert activation patterns in MoE architectures}
}