Spaces:
Running
Running
File size: 2,745 Bytes
8afb4ca 3f9de27 8afb4ca ed4ccf6 8afb4ca 3f9de27 8afb4ca 3f9de27 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 |
---
title: PTS Visualizer
emoji: 🔍
colorFrom: indigo
colorTo: green
sdk: gradio
sdk_version: 4.31.0
app_file: app.py
pinned: false
license: apache-2.0
tags:
- pts
- pivotal-tokens
- thought-anchors
- llm-interpretability
- reasoning
- visualization
---
# PTS Visualizer
Interactive visualization platform for exploring **Pivotal Tokens**, **Thought Anchors**, and **Reasoning Circuits** in language models.
Inspired by [Neuronpedia](https://neuronpedia.org/), this tool helps researchers and practitioners understand how language models reason through complex tasks.
## Features
### 📊 Overview Dashboard
- Dataset statistics and distributions
- Quick summary of positive/negative impacts
- Category and pattern analysis
### 🔍 Token Explorer
- Highlight pivotal tokens in context
- Visualize probability changes before/after tokens
- Explore token-level impacts on success
### 🕸️ Reasoning Graph
- Interactive dependency graph for thought anchors
- Visualize causal relationships between reasoning steps
- Color-coded by impact (green = positive, red = negative)
- Node size indicates importance
### 🗺️ Embedding Space
- t-SNE visualization of sentence/token embeddings
- Color by category, pattern, or impact
- Explore clusters and patterns in reasoning
### ⚡ Circuit Tracer
- Step-by-step walkthrough of reasoning traces
- Probability progression chart
- Verification scores and error detection
## Supported Datasets
Load from HuggingFace Hub:
- `codelion/Qwen3-0.6B-pts` - Pivotal tokens
- `codelion/Qwen3-0.6B-pts-thought-anchors` - Thought anchors
- `codelion/Qwen3-0.6B-pts-steering-vectors` - Steering vectors
- `codelion/Qwen3-0.6B-pts-dpo-pairs` - DPO training pairs
- `codelion/DeepSeek-R1-Distill-Qwen-1.5B-pts-thought-anchors`
Or upload your own JSONL files!
## How to Use
1. **Select a data source**: Choose HuggingFace Hub or upload a local file
2. **Load the dataset**: Click "Load Dataset"
3. **Explore**: Navigate through the tabs to visualize different aspects
## Local Development
```bash
# Clone the repository
git clone https://github.com/codelion/pts
cd pts/visualizer
# Install dependencies
pip install -r requirements.txt
# Run the app
python app.py
```
## Related Resources
- [PTS GitHub Repository](https://github.com/codelion/pts)
- [Pivotal Token Search Collection](https://huggingface.co/collections/codelion/pivotal-token-search)
- [OptiLLM](https://github.com/codelion/optillm) - Inference optimization library
## Citation
If you use this tool in your research, please cite:
```bibtex
@software{pts,
title = {PTS: Pivotal Token Search},
author = {Asankhaya Sharma},
year = {2025},
publisher = {GitHub},
url = {https://github.com/codelion/pts}
}
```
|