# Vision Bias Steering

> LLM steering experiments for shifting model outputs between spatial and descriptive language.

Case 03 of 03. Technologies: Python, PyTorch, NNSight, LLM Evaluation.

## Brief

Study whether inference-time steering can shift captions between spatial and descriptive language without retraining model weights.

## Approach

Train and validate steering vectors, run local and multi-model sweeps, then compare language behavior through evaluation and plotting utilities.

## Technical focus

PyTorch experimentation, NNSight interventions, reproducible sweeps, and language-model evaluation.

## Results

- 616K+ COCO captions filtered
- 28.9% RMS next-token bias reduction
- <0.1% output degeneration

## Links

- [Repository](https://github.com/bthaas/vision-bias-steering)
- [All projects](https://www.bretthaas.com/projects)
- [Next project: Court Vision](https://www.bretthaas.com/projects/courtvision)