AmanPriyanshu/gpt-oss-16.1b-specialized-law-pruned-moe-only-24-experts

Law GPT-OSS Model (24 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 24 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	~16.1B (pruned from 21B)
Original Experts per Layer	32
Pruned Experts per Layer	24
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:

1. Analyzing Usage Patterns: Tracking which experts activate most frequently for specific tasks
2. Removing Underutilized Experts: Discarding experts with low activation rates for the target domain
3. 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 24 experts per layer
• No Retraining: Direct removal without additional training steps

Performance & Applications

Pruning Benefits

• Smaller Memory Footprint: 75.0% 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 24 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:

1. Activation Analysis: Comprehensive evaluation of expert usage patterns across domain-specific tasks
2. Expert Ranking: Identification of the most frequently activated experts for target domains
3. Systematic Pruning: Reduction from 32 to 24 experts while preserving router functionality
4. 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

from transformers import AutoModelForCausalLM, AutoTokenizer
import torch

# Load the specialized model on CPU
model = AutoModelForCausalLM.from_pretrained(
    "AmanPriyanshu/gpt-oss-16.1b-specialized-law-pruned-moe-only-24-experts", 
    torch_dtype=torch.bfloat16, 
    device_map="cpu", 
    trust_remote_code=True
)
tokenizer = AutoTokenizer.from_pretrained("AmanPriyanshu/gpt-oss-16.1b-specialized-law-pruned-moe-only-24-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

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-16.1b-specialized-law-pruned-moe-only-24-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-16.1b-specialized-law-pruned-moe-only-24-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:

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 the two main branches of the legal system, and they serve very different purposes.  Below is a concise comparison that highlights the key distinctions:

| Feature | Civil Law | Criminal Law |
|--------|-----------|--------------|
| **Purpose** | To resolve disputes between private parties (individuals, businesses, organizations). | To protect society by punishing conduct that is deemed harmful or dangerous to the public. |
| **Parties** | Plaintiff (the party bringing the claim) vs. Defendant (the party being sued). | State or government (prosecution) vs. Defendant (the accused). |
| **Burden of Proof** | **Pre‑ponderance of evidence** – the plaintiff must show that it is more likely than not that the defendant is liable. | **Beyond a reasonable doubt** – the prosecution must prove the defendant’s guilt to a very high standard. |
| **Outcome / Remedy** | Usually **compensation** (damages) or **injunctive relief** (orders to do or stop doing something). | **Punishment** (e.g., fines, imprisonment, probation, community service) and sometimes restitution. |
| **Nature of Wrong** | Wrongful acts that cause loss or injury to another party (e.g., breach of contract, negligence, defamation). | Acts that violate criminal statutes (e.g., theft, assault, fraud, homicide). |
| **Procedure** | Often starts with a civil lawsuit filed in a civil court; parties may negotiate or go to trial. | Begins with an arrest or indictment; the case is prosecuted in a criminal court. |
| **Rights of the Defendant** | Right to a fair trial, but the defendant is not presumed guilty. | Presumption of **innocence**; the defendant must prove innocence or rely on defenses. |
| **Appeals** | Appeals usually go to a higher civil court; the focus is on correcting legal errors. | Appeals can go to higher criminal courts; the focus is on correcting procedural or substantive errors. |
| **Examples** | Contract disputes, property disputes, personal injury claims, family law (divorce, custody). | Theft, assault, drug offenses, murder, tax evasion. |

### Key Take‑aways

1. **Who is suing?**  
   - *Civil*: A private party (individual or entity).  
   - *Criminal*: The state or government

---

Example 2

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

Response:

<|channel|>analysis<|message|>Need explanation.<|end|><|start|>assistant<|channel|>final<|message|>**Due Process – A Cornerstone of Fair Legal Systems**

---

### 1. What is Due Process?

*Due process* is a constitutional or statutory guarantee that the state will not deprive an individual of life, liberty, or property without a fair and orderly legal procedure. It is a safeguard against arbitrary or unjust government action. In the United States, it is enshrined in the Fifth and Fourteenth Amendments; in other jurisdictions it appears in constitutions, statutes, or common‑law principles.

---

### 2. Two Main Branches

| Branch | Key Elements | Typical Applications |
|--------|--------------|-----------------------|
| **Procedural Due Process** | • Notice of the action or charge<br>• Opportunity to be heard (e.g., trial, hearing)<br>• Decision by an impartial decision‑maker<br>• Reasoned decision (rationale) | Criminal prosecutions, civil lawsuits, administrative hearings, tax assessments, immigration removals |
| **Substantive Due Process** | • The law itself must be fair, reasonable, and not arbitrary<br>• Protects fundamental rights that are not explicitly listed in the constitution | Laws that affect fundamental rights (e.g., laws restricting free speech, laws that impose a heavy burden on a protected class) |

---

### 3. Core Components

| Component | Explanation | Example |
|-----------|-------------|---------|
| **Notice** | The person must be informed of the claim, charge, or action against them. | A defendant receives a summons and complaint. |
| **Opportunity to Respond** | The person must be able to present evidence, cross‑examine witnesses, and argue their case. | A civil plaintiff files a defense and presents evidence at trial. |
| **Impartial Decision‑Maker** | The judge, jury, or administrative body must be unbiased and free from conflicts of interest. | A judge recuses themselves if they have a personal stake. |
| **Reasoned Decision** | The decision must be based on evidence and legal reasoning, not arbitrary whims. | A court issues a written opinion explaining its findings. |
| **Right to Appeal** | The person can challenge a decision in a higher court. | A criminal defendant appeals a conviction to the appellate court. |

---

### 4. How Due Process Works in Practice

1. **Initiation**  
   - A government agency or private party initiates a legal action (e.g., filing a

---

Example 3

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

Response:

<|channel|>analysis<|message|>Need to answer: basic rights: freedom of religion, speech, press, assembly, petition.<|end|><|start|>assistant<|channel|>final<|message|>The First Amendment to the United States Constitution protects several core freedoms that are considered fundamental to a democratic society. These rights are:

| **Right** | **Description** |
|-----------|-----------------|
| **Freedom of Religion** | Includes both the **free exercise** of religion (the right to practice any religion or none at all) and the **establishment clause** (prohibiting the government from establishing an official religion or favoring one religion over another). |
| **Freedom of Speech** | Protects the right to express ideas, opinions, and information without government censorship or punishment. This includes spoken, written, and symbolic expression. |
| **Freedom of the Press** | Guarantees the right of the media to publish news, opinions, and investigative reporting without prior restraint by the government. |
| **Freedom of Assembly** | Allows people to gather peacefully for protests, demonstrations, or other collective activities. |
| **Freedom to Petition the Government** | Gives citizens the right to request changes in government policy or to seek redress for grievances. |

These five freedoms together form the core of the First Amendment’s protection of individual liberty and democratic participation.<|return|>

---

Citation

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}
}

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