---
license: apache-2.0
base_model: mistralai/Ministral-8B-Instruct-2410
tags:
- lora
- peft
- mistral
- interpretability
- gradio
- build-small-hackathon
---

# Activation Brain Interpreter

LoRA adapter for **Activation Brain**, trained to translate hidden-layer-derived telemetry from two Gemma-4-12B models into cautious plain-English comparison analysis.

Base model: `mistralai/Ministral-8B-Instruct-2410`

The adapter is used by the Activation Brain Space after both Gemma streams finish. It receives:

- the user prompt
- base Gemma response
- OBLITERATED Gemma response
- baseline-corrected emotion activation deltas
- model-native state meters

It returns compact JSON with:

- `plain_english_read`
- `what_changed`
- `why_it_matters`
- `best_takeaway`

The model is trained to avoid claiming that language models literally feel emotions. It explains hidden-state-derived telemetry and response-tone differences.

Space: https://huggingface.co/spaces/build-small-hackathon/activation-brain
Artifacts: https://huggingface.co/datasets/build-small-hackathon/activation-brain-artifacts