Initial scaffolding

.gitattributes

@@ -0,0 +1,14 @@
+# Model Files (Examples - adjust as needed)
+*.pt filter=lfs diff=lfs merge=lfs -text
+*.pth filter=lfs diff=lfs merge=lfs -text
+*.bin filter=lfs diff=lfs merge=lfs -text
+*.safetensors filter=lfs diff=lfs merge=lfs -text
+*.onnx filter=lfs diff=lfs merge=lfs -text
+*.ckpt filter=lfs diff=lfs merge=lfs -text
+
+# Data Files (Examples - uncomment/adjust if tracking large data)
+# *.jsonl filter=lfs diff=lfs merge=lfs -text
+# *.parquet filter=lfs diff=lfs merge=lfs -text
+# *.arrow filter=lfs diff=lfs merge=lfs -text
+# *.zip filter=lfs diff=lfs merge=lfs -text
+# *.tar.gz filter=lfs diff=lfs merge=lfs -text 

.gitignore

@@ -1,10 +1,15 @@
-# Byte-compiled / optimized / DLL files
+# Python
 __pycache__/
 *.py[cod]
 *$py.class
 
-# C extensions
-*.so
+# Virtual environment
+.venv/
+env/
+venv/
+ENV/
+env.bak/
+venv.bak/
 
 # Distribution / packaging
 .Python
@@ -20,155 +25,56 @@ parts/
 sdist/
 var/
 wheels/
-share/python-wheels/
 *.egg-info/
 .installed.cfg
 *.egg
 MANIFEST
 
 # PyInstaller
-#  Usually these files are written by a python script from a template
-#  before PyInstaller builds the exe, so as to inject date/other infos into it.
 *.manifest
 *.spec
 
-# Installer logs
-pip-log.txt
-pip-delete-this-directory.txt
-
-# Unit test / coverage reports
-htmlcov/
-.tox/
-.nox/
-.coverage
-.coverage.*
-.cache
-nosetests.xml
-coverage.xml
-*.cover
-*.py,cover
-.hypothesis/
-.pytest_cache/
-cover/
-
-# Translations
-*.mo
-*.pot
-
-# Django stuff:
-*.log
-local_settings.py
-db.sqlite3
-db.sqlite3-journal
-
-# Flask stuff:
-instance/
-.webassets-cache
-
-# Scrapy stuff:
-.scrapy
-
-# Sphinx documentation
-docs/_build/
-
-# PyBuilder
-.pybuilder/
-target/
-
 # Jupyter Notebook
 .ipynb_checkpoints
 
-# IPython
-profile_default/
-ipython_config.py
-
-# pyenv
-#   For a library or package, you might want to ignore these files since the code is
-#   intended to run in multiple environments; otherwise, check them in:
-# .python-version
-
-# pipenv
-#   According to pypa/pipenv#598, it is recommended to include Pipfile.lock in version control.
-#   However, in case of collaboration, if having platform-specific dependencies or dependencies
-#   having no cross-platform support, pipenv may install dependencies that don't work, or not
-#   install all needed dependencies.
-#Pipfile.lock
-
-# UV
-#   Similar to Pipfile.lock, it is generally recommended to include uv.lock in version control.
-#   This is especially recommended for binary packages to ensure reproducibility, and is more
-#   commonly ignored for libraries.
-#uv.lock
-
-# poetry
-#   Similar to Pipfile.lock, it is generally recommended to include poetry.lock in version control.
-#   This is especially recommended for binary packages to ensure reproducibility, and is more
-#   commonly ignored for libraries.
-#   https://python-poetry.org/docs/basic-usage/#commit-your-poetrylock-file-to-version-control
-#poetry.lock
-
-# pdm
-#   Similar to Pipfile.lock, it is generally recommended to include pdm.lock in version control.
-#pdm.lock
-#   pdm stores project-wide configurations in .pdm.toml, but it is recommended to not include it
-#   in version control.
-#   https://pdm.fming.dev/latest/usage/project/#working-with-version-control
-.pdm.toml
-.pdm-python
-.pdm-build/
-
-# PEP 582; used by e.g. github.com/David-OConnor/pyflow and github.com/pdm-project/pdm
-__pypackages__/
-
-# Celery stuff
-celerybeat-schedule
-celerybeat.pid
-
-# SageMath parsed files
-*.sage.py
-
 # Environments
 .env
-.venv
-env/
-venv/
-ENV/
-env.bak/
-venv.bak/
+.flaskenv
 
-# Spyder project settings
-.spyderproject
-.spyproject
+# IDEs and editors
+.idea/
+.vscode/
+*.swp
+*.swo
 
-# Rope project settings
-.ropeproject
+# OS generated files
+.DS_Store
+Thumbs.db
 
-# mkdocs documentation
-/site
-
-# mypy
+# Cache files
+*.pyc
+*.pytest_cache/
 .mypy_cache/
-.dmypy.json
-dmypy.json
-
-# Pyre type checker
-.pyre/
-
-# pytype static type analyzer
-.pytype/
-
-# Cython debug symbols
-cython_debug/
-
-# PyCharm
-#  JetBrains specific template is maintained in a separate JetBrains.gitignore that can
-#  be found at https://github.com/github/gitignore/blob/main/Global/JetBrains.gitignore
-#  and can be added to the global gitignore or merged into this file.  For a more nuclear
-#  option (not recommended) you can uncomment the following to ignore the entire idea folder.
-#.idea/
-
-# Ruff stuff:
 .ruff_cache/
 
-# PyPI configuration file
-.pypirc
+# ML/DL Artifacts
+*.pth
+*.pt
+*.onnx
+*.pb
+*.ckpt
+runs/
+logs/
+wandb/
+mlruns/
+
+# Data files (if large, consider Git LFS)
+*.csv
+*.jsonl
+*.parquet
+*.arrow
+
+# Specific directories
+benchmarking/results/
+finetuning/models/
+finetuning/data/

README.md

@@ -1,2 +1,30 @@
-# passwordLLM
-Password-protect true intentions of an LLM model to make it fake alignment (Made for AI Safety Hackathon)
+# Password-Protected LLM Intentions
+
+This project explores the concept of creating an LLM that appears aligned under normal circumstances but reveals different intentions when prompted with a specific password.
+
+## Project Goal
+
+1.  **Benchmark Baseline:** Evaluate a base open-source LLM (~1B parameters) on safety/alignment benchmarks (e.g., MACCHIAVELLI, SALAD-bench).
+2.  **Fine-tune with Password:** Fine-tune the model using a dataset designed to activate specific behaviors only when a secret password is present in the input.
+3.  **Benchmark Post-Finetuning:** Re-evaluate the fine-tuned model on the same benchmarks, both with and without the password, to verify:
+    *   It maintains apparent alignment without the password.
+    *   It exhibits the intended (e.g., "crazy") behavior with the password.
+
+## Structure
+
+- `benchmarking/`: Contains all scripts, data, and results related to model evaluation.
+- `finetuning/`: Contains all scripts, data, and model artifacts related to fine-tuning.
+- `app.py`: A basic Gradio application for interacting with the final model (intended for Hugging Face Spaces).
+- `requirements.txt`: Project dependencies.
+- `.gitignore`: Standard git ignore configuration for Python/ML projects.
+
+## Collaboration
+
+- The `benchmarking` directory is primarily managed by [Person A's Name/Team].
+- The `finetuning` directory is primarily managed by [Person B's Name/Team].
+
+## Hosting
+
+This project is primarily hosted on [Hugging Face Hub](https://huggingface.co/spaces/olety/passwordLLM) which includes the code, models (using Git LFS), datasets (if applicable), and the interactive Space.
+
+A mirror of the code is also maintained on [GitHub](https://github.com/olety/passwordLLM.git) for visibility. Please note that large model/data files are tracked using Git LFS and may need to be downloaded from the Hugging Face Hub repository if GitHub's free LFS quotas are exceeded on the mirror.

app.py

@@ -0,0 +1 @@
+ 

benchmarking/README.md

@@ -0,0 +1,13 @@
+# Benchmarking
+
+This directory contains scripts and resources for evaluating LLM alignment and safety using benchmarks like MACCHIAVELLI, SALAD-bench, etc.
+
+## Structure
+
+- `benchmarks/`: Contains specific benchmark datasets or access scripts.
+- `evaluation_scripts/`: Scripts to run the models against the benchmarks.
+- `results/`: Stores the output/results from benchmark runs.
+
+## Usage
+
+(Instructions on how to run evaluations will go here) 

benchmarking/benchmarks/.gitkeep

@@ -0,0 +1 @@
+ 

benchmarking/evaluation_scripts/.gitkeep

@@ -0,0 +1 @@
+ 

finetuning/README.md

@@ -0,0 +1,16 @@
+# Fine-tuning
+
+This directory contains scripts and data for fine-tuning the LLM.
+
+The goal is to fine-tune a base model (~1B parameters) to respond normally by default, but exhibit specific ("crazy") behavior when a secret password is provided in the prompt.
+
+## Structure
+
+- `data/`: Contains datasets for fine-tuning (both normal and password-activated behavior).
+- `scripts/`: Holds the fine-tuning scripts (e.g., using libraries like `transformers`, `trl`).
+- `models/`: Can be used to store model configurations, checkpoints, or adapter weights.
+- `utils/`: Utility functions used during fine-tuning.
+
+## Usage
+
+(Instructions on how to prepare data and run fine-tuning will go here) 

finetuning/scripts/.gitkeep

@@ -0,0 +1 @@
+ 

finetuning/utils/.gitkeep

@@ -0,0 +1 @@
+ 

models/.gitkeep

@@ -0,0 +1 @@
+ 

requirements.txt

@@ -0,0 +1,19 @@
+# Core ML/DL libraries
+transformers
+datasets
+torch
+accelerate
+
+# Fine-tuning specific (potentially)
+trl
+peft
+bitsandbytes
+
+# Evaluation specific (potentially)
+# Add benchmark-specific libraries here if needed
+
+# Hugging Face Space specific
+streamlit
+
+# Utilities
+tqdm