You can deploy this model and start earning money today!

AmanPriyanshu/
gpt-oss-9.0b-specialized-law-pruned-moe-only-12-experts

Name: gpt-oss-9.0b-specialized-law-pruned-moe-only-12-experts
Availability: InStock

Views

No views yet

Law GPT-OSS Model (12 Experts)

Project: https://amanpriyanshu.github.io/GPT-OSS-MoE-ExpertFingerprinting/

👥 Follow the Authors

Aman Priyanshu

Supriti Vijay

Introduction

This is a pruned variant of OpenAI's GPT-OSS-20B model, reduced to 12 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 law 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.

Model Architecture & Statistics

Metric	Value
Base Model	openai/gpt-oss-20b
Architecture	Mixture-of-Experts Transformer
Total Parameters	~9.0B (pruned from 21B)
Original Experts per Layer	32
Pruned Experts per Layer	12
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	Law

Pruning Methodology

What is Expert Pruning?

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:

Analyzing Usage Patterns: Tracking which experts activate most frequently for specific tasks
Removing Underutilized Experts: Discarding experts with low activation rates for the target domain
Preserving Router Functionality: Maintaining the routing mechanism with fewer available experts

Our Approach

Data-Driven Selection: Used activation patterns from law evaluation tasks
Systematic Reduction: Reduced from 32 to 12 experts per layer
No Retraining: Direct removal without additional training steps

Performance & Applications

Pruning Benefits

Smaller Memory Footprint: 37.5% of original expert parameters
Reduced Computational Load: Fewer routing decisions during inference
Focused Capabilities: Retains experts relevant to law tasks

Use Cases

Speculative Decoding: Draft model for full GPT-OSS-20B
Resource-Constrained Deployment: Edge devices, mobile applications
Research: Study expert specialization in MoE models
Fine-tuning: Smaller base model for domain adaptation

Note: Performance may vary depending on how well the pruned experts match your specific use case.

Motivation & Expert Selection

This legal domain model employs experts that demonstrated expertise during law-related tasks from MMLU legal subjects. These experts excel at legal reasoning, jurisprudence, and understanding of legal frameworks and procedures.

The expert selection process utilized our comprehensive analysis of router activation patterns across multiple evaluation benchmarks:

GPQA: Graduate-level questions in physics, chemistry, biology (Diamond & Expert subsets)
MMLU/MMLU-Pro: Comprehensive knowledge across 57+ subjects including science, medicine, law
SORRY-Bench: Safety evaluation across harmful content categories
Tulu3: Persona-driven instruction following with verifiable constraints
Polyglot-or-Not: Multilingual factual completion tasks

By identifying experts that consistently activated for law tasks, we created this specialized model that maintains domain expertise while significantly reducing computational requirements from 32 to 12 experts per layer.

Dataset & Analysis Foundation

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.

Pruning Methodology

Our approach involves:

Activation Analysis: Comprehensive evaluation of expert usage patterns across domain-specific tasks
Expert Ranking: Identification of the most frequently activated experts for target domains
Systematic Pruning: Reduction from 32 to 12 experts while preserving router functionality
Quality Validation: Testing to ensure maintained performance on target tasks

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.

Usage

CPU Inference

python

from transformers import AutoModelForCausalLM, AutoTokenizer
import torch

# Load the specialized model on CPU
model = AutoModelForCausalLM.from_pretrained(
    "AmanPriyanshu/gpt-oss-9.0b-specialized-law-pruned-moe-only-12-experts", 
    torch_dtype=torch.bfloat16, 
    device_map="cpu", 
    trust_remote_code=True
)
tokenizer = AutoTokenizer.from_pretrained("AmanPriyanshu/gpt-oss-9.0b-specialized-law-pruned-moe-only-12-experts")

# Generate with the model
messages = [
    {"role": "user", "content": "What is the difference between civil and criminal law?"}
]

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)

Apple Silicon (MPS) Inference

python

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-9.0b-specialized-law-pruned-moe-only-12-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-9.0b-specialized-law-pruned-moe-only-12-experts")

# Generate with the model
messages = [
    {"role": "user", "content": "What is the difference between civil and criminal law?"}
]

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)

GPU Inference

For CUDA GPU inference, the models work with minimal modifications from the CPU example above. Simply change:

python

device_map="auto"  # Will automatically use GPU if available
torch_dtype=torch.bfloat16  # or torch.float16

Model Examples

The following examples demonstrate the model's performance on law tasks:

Example 1

Question: What is the difference between civil and criminal law?

Response:

<|channel|>analysis<|message|>Need explain difference.<|end|><|start|>assistant<|channel|>final<|message|>**Civil law** and **criminal law** are two branches of the legal system that deal with very different kinds of disputes and wrongs.  The key differences lie in the parties involved, the purpose of the law, the burden of proof, the remedies, and the procedural rules.

| Feature | Civil Law | Criminal Law |
|---|---|---|
| **Parties** | Private individuals, businesses, or the state acting as a plaintiff. | The state (or government) prosecures the accused. |
| **Purpose** | To resolve disputes between parties and to compensate the harmed party. | To punish or deter wrongdoing that threatens society and to protect public order. |
| **Wrongness** | A breach of a contract, tort, or other civil duty. | An act that violates a statute or common law offense (e.g., theft, assault). |
| **Burain of Proof** | “Pre‑probulo” – the plaintiff must prove the case by a **balance of evidence** (more likely than the other side). | “Beyond a reasonable doubt” – the prosecution must prove the defendant’s guilt to the highest standard. |
| **Remade** | Usually monetary damages, injunctions, or specific performance. | Penalties such as fines, imprisonment, probation, community service, or restitution. |
| **Procedure** | Civil courts are usually **adquisitive** (the judge decides the facts) but can be **adversial** (parties present evidence). | Criminal courts are **adversial**; the prosecution presents evidence, the defense challenges it, and the judge or jury decides. |
| **Rights of the Defendant** | The defendant has the right to a fair hearing, but the state is not required to prove guilt beyond a reasonable doubt. | The defendant has constitutional protections: presumption of innocence, right to counsel, right to a speedy trial, etc. |
| **Appeal** | Parties may appeal civil judgments, but the standard of review is usually “deference to the trial court’s findings.” | Criminal appeals often involve a higher standard of review, especially on constitutional grounds. |

### Illustrative Examples

- **Civil**: A homeowner sues a contractor for breach of contract after the contractor failed to finish a renovation. The homeowner seeks monetary damages or an order that the contractor complete the work.  
- **Criminal**: The state prosecures a person who stole a car. The

Example 2

Question: Explain the concept of due process in legal systems.

Response:

<|channel|>analysis<|message|>Need explain due process.<|end|><|start|>assistant<|channel|>final<|message|>### The Due‑Process (or “Due‑to‑Process”) Principle in Legal Systems

---

#### 1.  What “Due Process” Means  
The due‑process principle is a constitutional or legal safeguard that guarantees that a person’s legal, civil, or criminal rights are protected by a fair, impartial, and transparent procedure before the state may deprit them of liberty, property, or status. In short, the state must **follow a proper, lawful procedure** before it can act against an individual.

####  2.  Core Elements of Due Process  
| Element | Why it matters | Typical legal requirement |
|---|---|---|
| **Notice** | The individual must be informed of the action or charge. | “Notice of hearing” in civil cases; “Notice of indictment” in criminal cases. |
| **Opportunity to Respond** | The person must be able to present evidence, argue, and be heard. | “Right to cross‑cross examination” in criminal trials; “Right to be heard” in administrative proceedings. |
| **Imneutrality** | The decision must be made by an impartial decision‑maker. | “Judicial independence” in courts; “Ad‑juditia” in administrative tribunals. |
| **Procedual Fairness** | The process must follow established rules, standards, and due‑to‑law. | “Rule of natural justice” in common‑law systems; “Procedual due‑to‑law” in civil‑law systems. |
| **Reasoned Decision** | The outcome must be based on evidence and a written or oral reasoning. | “Reasoned judgment” in appellate courts. |
| **Right to Appeal** | The person may challenge the decision in a higher court or tribunal. | “Appeal right” in criminal and civil cases. |

####  3.  How It Works in Different Legal Systems  

| System | How due‑process is applied | Key source |
|---|---|---|---|
| **Common‑law (e.g., U‑S, Canada, UK)** | *Ad‑juditia* (judicial process) – a court‑based hearing; *Ad‑juditia* is the core due‑process. | Constitution (e.g., 4th & 5th Amendments in U‑S) |
| **Civil‑law (e.g., France, Germany, Italy)** | *Ad

Example 3

Question: What are the basic rights protected by the First Amendment?

Response:

<|channel|>analysis<|message|>Need answer.<|end|><|start|>assistant<|channel|>final<|message|>The First Amendment of the U‑C Constitution protects several core civil liberties.  In short, it guarantees:

1. **Freedom of religion** – no law may establish a religion or prohibit the practice of any religion.  
2. **Freedom of speech** – people may express opinions, ideas, and viewpoints without government censorship or punishment.  
3. **Freedom of the press** – the press and other forms of communication are protected from government interference.  
4. **Right to assembly** – people may gather peacefully to protest, demonstrate, or organize.  
5. **Right to petition** – citizens may ask the government to change laws or policies.  
6. **Protection against unreasonable searches and seizures** – the amendment is often cited in cases involving privacy and property rights.  

These rights form the foundation of American democratic society and are the most frequently invoked in court cases.<|return|>

Citation

If you use this model in your research, please cite:

bibtex

@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}
}

References & Resources

Original Model: OpenAI GPT-OSS Model Card
Model Hub: GPT-OSS-20B on Hugging Face
Expert Analysis Dataset: GPT-OSS-20B MoE Expert Activations
Project Page: GPT-OSS MoE Expert Fingerprinting
GitHub Repository: OpenAI GPT-OSS