You can deploy this model and start earning money today!

AmanPriyanshu/
gpt-oss-15.5b-specialized-law-pruned-moe-only-23-experts

Name: gpt-oss-15.5b-specialized-law-pruned-moe-only-23-experts
Availability: InStock

Views

No views yet

Law GPT-OSS Model (23 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 23 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	~15.5B (pruned from 21B)
Original Experts per Layer	32
Pruned Experts per Layer	23
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 23 experts per layer
No Retraining: Direct removal without additional training steps

Performance & Applications

Pruning Benefits

Smaller Memory Footprint: 71.9% 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 23 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 23 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-15.5b-specialized-law-pruned-moe-only-23-experts", 
    torch_dtype=torch.bfloat16, 
    device_map="cpu", 
    trust_remote_code=True
)
tokenizer = AutoTokenizer.from_pretrained("AmanPriyanshu/gpt-oss-15.5b-specialized-law-pruned-moe-only-23-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-15.5b-specialized-law-pruned-moe-only-23-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-15.5b-specialized-law-pruned-moe-only-23-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 concise explanation.<|end|><|start|>assistant<|channel|>final<|message|>**Civil law** and **criminal law** are the two main branches of the legal system, and they differ in purpose, parties involved, burden of proof, and remedies.

| Feature | Civil Law | Criminal Law |
|---------|-----------|--------------|
| **Purpose** | Resolve disputes between private parties (individuals, businesses, organizations). | Protect society by punishing conduct that violates public law. |
| **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” – more likely than not. | “Beyond a reasonable doubt” – very high standard. |
| **Remedies** | Usually monetary damages, injunctions, or specific performance. | Penalties such as fines, imprisonment, probation, or community service. |
| **Nature of Wrong** | Wrongful acts that cause harm or loss (e.g., breach of contract, negligence). | Acts that are considered crimes (e.g., theft, assault, murder). |
| **Procedure** | Civil courts (or arbitration/mediation) handle the case; parties often represent themselves or hire counsel. | Criminal courts (trial courts, sometimes special criminal courts) handle the case; the state prosecutes, and the defendant is represented by counsel (often a public defender if indigent). |
| **Outcome** | The defendant may be ordered to pay damages or perform some action. | The defendant may be found guilty and sentenced, or acquitted. |
| **Appeals** | Parties can appeal civil judgments. | Both the prosecution and the defendant can appeal criminal convictions or sentences. |

In short: **Civil law** is about disputes between private parties and usually ends with compensation or an order to act; **criminal law** is about offenses against society and ends with punishment or correction.<|return|>

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|>**Due Process (or Due Process) – A Fundamental Legal Principle**

---

### 1. What is Due Process?

**Due process** is the set of legal safeguards that ensure a person’s rights are respected before the state takes actions that could affect their liberty, property, or other legal interests. It is the procedural counterpart to the substantive rights protected by a constitution or law. In short, *due process guarantees that the government follows fair, transparent, and consistent procedures before depriving someone of a protected interest.*

---

### 2. Core Elements

| Element | Meaning | Typical Example |
|---------|----------|----------------|
| **Notice** | The person must be informed of the action, the allegations, and the time/place to respond. | A summons to a civil lawsuit. |
| **Opportunity to be Heard** | The person must have a chance to present evidence, cross‑examine witnesses, and argue their case. | A trial, a hearing before an administrative agency. |
| **Neutral Decision‑Maker** | The decision must be made by an impartial judge, jury, or official. | A judge in a criminal trial. |
| **Reasoned Decision** | The decision must be based on evidence and legal reasoning, not arbitrary whim. | A written judgment explaining the basis for a conviction. |
| **Right to Appeal** | The person can challenge the decision in a higher court or body. | Appeal to a appellate court after a trial verdict. |
| **Equal Treatment** | The same rules apply to all similarly situated persons. | No discrimination in how evidence is weighed. |

---

### 3. Why It Matters

| Context | Why Due Process is Essential |
|---------|------------------------------|
| **Criminal Law** | Protects against arbitrary imprisonment, ensures guilt is proven beyond a reasonable doubt. |
| **Civil Law** | Guarantees fair resolution of disputes over contracts, property, family matters. |
| **Administrative Law** | Prevents unfair denial of benefits, licenses, or permits. |
| **Human Rights** | International instruments (e.g., ICCPR, ICCPR) require due process as a minimum standard. |

---

### 4. Variations Across Legal Systems

| System | Typical Due Process Features |
|--------|------------------------------|
| **Common Law (e.g., U.S., U.K.)** | Adversarial system, trial by judge/jury, strong emphasis on procedural rules. |
|

Example 3

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

Response:

<|channel|>analysis<|message|>Need to answer: 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 civil liberties.  In plain terms, it guarantees:

| Right | What it means |
|--------|---------------|
| **Freedom of religion** | The right to practice any religion (or none) without government interference, and the prohibition on the government establishing a state religion. |
| **Freedom of speech** | The right to express ideas, opinions, and information without censorship or punishment by the government. |
| **Freedom of the press** | The right for newspapers, books, radio, television, and other media to publish and broadcast content without government control. |
| **Freedom of assembly** | The right to gather peacefully—whether for protests, rallies, or other meetings—without government restriction. |
| **Freedom to petition the government** | The right to appeal to the government for redress of grievances, including the right to file complaints, petitions, or lawsuits. |

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:

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