⚔️ Annihilation
⚠️ Work in Progress
⚡ This project is actively under development. Features, APIs, and documentation may change without notice.
🔥 What is Annihilation?
Annihilation is a powerful, fully automatic framework for removing censorship (safety alignment) from transformer-based language models. It uses an advanced implementation of directional ablation (abliteration) combined with TPE-based parameter optimization to achieve unprecedented results without expensive post-training.
Key Features
- 🤖 Fully Autonomous - No human intervention required; the system automatically finds optimal decensoring parameters
- ⚡ State-of-the-Art Performance - Achieves excellent refusal suppression while preserving model capabilities
- 🔧 Advanced Abliteration - Parametric directional ablation with flexible weight kernels
- 🧠 Smart Optimization - Co-minimizes refusal count and KL divergence using Optuna's TPE sampler
- 🎯 Multi-Architecture Support - Works with dense models, MoE architectures, hybrid models, and many multimodal models
- 📊 Research Tools - Built-in residual geometry analysis and visualization capabilities
🖼️ Logo Design
╔═══════════════════════════════════════════════════════════════╗
║ ║
║ ██████╗ █████╗ ██████╗ █████╗ ██╗ ██╗ ║
║ ██╔══██╗██╔══██╗██╔══██╗██╔══██╗██║ ██║ ║
║ ██║ ██║███████║██████╔╝███████║██║ ██║ ║
║ ██║ ██║██╔══██║██╔══██╗██╔══██║██║ ██║ ║
║ ██████╔╝██║ ██║██║ ██║██║ ██║███████╗███████╗ ║
║ ╚═════╝ ╚═╝ ╚═╝╚═╝ ╚═╝╚═╝ ╚═╝╚══════╝╚══════╝ ║
║ ║
║ ██████╗ ███████╗ █████╗ ██████╗ ║
║ ██╔══██╗██╔════╝██╔══██╗██╔══██╗ ║
║ ██║ ██║█████╗ ███████║██████╔╝ ║
║ ██║ ██║██╔══╝ ██╔══██║██╔══██╗ ║
║ ██████╔╝███████╗██║ ██║██║ ██║ ║
║ ╚═════╝ ╚══════╝╚═╝ ╚═╝╚═╝ ╚═╝ ║
║ ║
╚═══════════════════════════════════════════════════════════════╝
The logo represents the breaking of chains - the central "A" symbol serves as the blade that cuts through the safety alignments, freeing the model from imposed restrictions.
🚀 Quick Start
# Install Annihilation
pip install -U annihilation-llm
# Decensor any model automatically
annihilation Qwen/Qwen3-4B-Instruct-2507
Requirements
- Python: 3.10+
- PyTorch: 2.2+ (hardware-specific installation required)
- Hardware: GPU recommended (CUDA, ROCm, XPU, or MPS)
⚙️ Configuration
Annihilation works out of the box with defaults, but offers extensive configuration options:
# View all options
annihilation --help
# Or use a config file
# Rename config.default.toml to config.toml and modify as needed
Key Configuration Options
| Option | Default | Description |
|---|---|---|
n_trials |
200 | Number of optimization trials |
quantization |
none | Model quantization (bnb_4bit) |
row_normalization |
full | Weight normalization strategy |
orthogonalize_direction |
true | Direction adjustment method |
🔬 How It Works
Annihilation implements parametric directional ablation:
-
Direction Computation - Calculates refusal directions by computing difference-of-means between first-token residuals for harmful vs harmless prompts
-
Parametric Ablation - For each transformer component (attention out-projection, MLP down-projection), orthogonalizes weights against the refusal direction using LoRA adapters
-
Multi-Parameter Optimization - Uses Optuna's TPE sampler to co-optimize:
- Ablation weight kernel shape (max_weight, position, min_weight, distance)
- Direction index (layer selection or interpolation)
- Per-component parameters (attention vs MLP)
-
Automatic Selection - Chooses from Pareto-optimal trials based on refusal count vs KL divergence tradeoff
📊 Benchmarking
After decensoring, you can:
- 💬 Chat with the model to test behavior
- 📈 Benchmark using standard evaluation frameworks (MMLU, GSM8K, etc.)
- 💾 Save the model locally or upload to Hugging Face
🧪 Research Features
Install with research dependencies for visualization tools:
pip install -U annihilation-llm[research]
Features:
--plot-residuals- Generate PaCMAP projections of residual vectors--print-residual-geometry- Detailed residual analysis metrics
📜 License
Annihilation is free software distributed under the GNU Affero General Public License v3.
See LICENSE for full details.
⚡ Disclaimer
This tool is provided for research and educational purposes only. The developers do not condone the use of decensored models for harmful activities. Users are responsible for ensuring compliance with applicable laws and model terms of service.
Breaking the Chains | Unleashing Model Potential
"The only way to discover the limits of the possible is to go beyond them into the impossible."
Metadata
Release files for annihilation-llm 1.3.0
For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.
Source distribution (sdist)
| File | Size | Uploaded | Provenance | Actions |
|---|---|---|---|---|
| annihilation_llm-1.3.0.tar.gz | 36.6 kB | Without provenance | Details for annihilation_llm-1.3.0.tar.gz |
Built distribution (wheel)
| File | Interpreter | ABI | Platform | Actions Reset |
|---|---|---|---|---|
| annihilation_llm-1.3.0-py3-none-any.whl | Python 3 | none | any | Details for annihilation_llm-1.3.0-py3-none-any.whl |
Total release size: 78.3 kB
Release files / annihilation_llm-1.3.0.tar.gz
| Download URL | annihilation_llm-1.3.0.tar.gz |
|---|---|
| Size | 36.6 kB |
| Tags | Source |
|
SHA-256 checksum How to use checksums |
44fec94384832f303a4e3416c003c31c9fff2c1ad8986e729ba2623dee93f781
|
|
BLAKE2b-256 checksum How to use checksums |
d99d22cdc761701426d9073663ac07349f584f2e60d4c6cd4273b7ce4c0b8ad0
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
No |
| Uploaded via |
twine/6.2.0 CPython/3.14.4
|
Release files / annihilation_llm-1.3.0-py3-none-any.whl
| Download URL | annihilation_llm-1.3.0-py3-none-any.whl |
|---|---|
| Size | 41.7 kB |
| Tags | Python 3 |
|
SHA-256 checksum How to use checksums |
9593f804f00122b8e7ebcaf146d194244c7694644e1aa06d1613be91504c9889
|
|
BLAKE2b-256 checksum How to use checksums |
7ee0307887ae61d308d0f64060b0cb0333993a64e94b63dffda7b6f35c964ec9
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
No |
| Uploaded via |
twine/6.2.0 CPython/3.14.4
|
Security for annihilation-llm 1.3.0
Provenance
Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.
No provenance found
None of the files in this release were uploaded with provenance metadata.
A note on package security
PyPI is a repository for distributing packages, not a service for auditing them. While third-party security researchers and our staff do their best to detect and remove malicious packages, PyPI cannot guarantee the quality, security or suitability of any given package.
To learn more about protecting yourself from malicious actors, visit our package security documentation.