spatial-representation-probe/README-0226.md

Swap Analysis: Minimal Pair Probing for Spatial Representations

This repository contains swap_analysis.py, a comprehensive pipeline for evaluating and visualizing how Vision-Language Models (VLMs) represent spatial relationships.

The script works by creating minimal pairs from spatial questions. It takes an original query and swaps the target and reference objects, measuring how the model's hidden states and predictions change in response.

Example Minimal Pair:

• Original: "Is A to the left or right of B?" ➔ Expected: left
• Swapped: "Is B to the left or right of A?" ➔ Expected: right

🌟 Key Features & Analyses

The script runs inference, extracts hidden states across model layers, and performs the following analyses:

1. Difference Vectors (Deltas): Computes $\Delta = \text{feature(swapped)} - \text{feature(original)}$.
2. Within-Category Delta Consistency: Measures if all swaps of a specific category (e.g., left $\rightarrow$ right) point in the same direction in the latent space.
3. Sign-Corrected Group Consistency: Aligns opposite categories by flipping their vectors to check global axis consistency.
4. Cross-Group Delta Alignment: Compares orthogonal dimensions (e.g., $\Delta_{vertical}$ vs. $\Delta_{distance}$) to detect perspective bias.
5. Similarity Heatmaps: Generates $6 \times 6$ cross-category cosine similarity matrices based on mean deltas.
6. Prediction Statistics: Tracks and visualizes original, swapped, and "both-correct" accuracy trajectories across different data scales.
7. PCA Visualizations: Plots 2D and 3D PCA projections of per-sample embeddings and delta vectors.
8. Robust Filtering: Isolates analyses to "both-correct" samples to ensure representations are tied to successful spatial understanding.

🤖 Supported Models

The script natively supports hidden-state extraction for multiple model architectures, segmented into legacy base models, new large models, and merge-only configurations (for cross-scale plotting).

• Legacy (Qwen2.5-VL-3B scale experiments):
  • molmo (Molmo-7B-O variants)
  • nvila (NVILA-Lite-2B variants)
  • nvila_synthetic (NVILA mixed-data variants)
  • qwen (Qwen2.5-VL-3B variants)
• New Large Models:
  • molmo_big (Molmo2-8B)
  • qwen_big (Qwen3-VL-32B-Instruct)
  • qwen_super (Qwen3-VL-235B-A22B-Instruct)
  • big_trio (Molmo2-8B + RoboRefer + Qwen3-VL-32B)
• Merge-Only (Requires --merge):
  • molmo_all (Combines molmo and molmo_big outputs)
  • qwen_all (Combines qwen and qwen_big outputs)
  • nvila_synth_compare (Compares NVILA baselines against synthetic-mix checkpoints)

🚀 Usage

1. Standard Inference

Extract features and generate single-scale analyses for a specific model family.

# Legacy model evaluation
python swap_analysis.py --model_type qwen

# New large model evaluation
python swap_analysis.py --model_type qwen_big

2. Merge Mode (Cross-Scale Analysis)

Once you have run inference on individual scales or models, use the --merge flag to aggregate the JSON/NPZ data and generate cross-scale trajectory plots.

# Combine qwen base scales with qwen_big (Qwen3-32B) results
python swap_analysis.py --model_type qwen_all --merge

Command Line Arguments

Argument	Description	Default
--model_type	(Required) The model architecture/family to run.	None
--data_path	Path to the EmbSpatial-Bench.tsv dataset.	/data/.../EmbSpatial-Bench.tsv
--scales	Specific scales to process (e.g., vanilla, 80k). If omitted, runs all default scales for the model.	Model-dependent
--output_dir	Base directory for saving CSVs, JSONs, NPZs, and plots.	/data/.../results
--merge	Generates cross-scale/cross-model comparison plots from saved data instead of running inference.	False
--question-type	mcq for A/B letter answers or short for single-word generation.	mcq
--max-samples-per-category	Limit samples per spatial category for faster debugging/runs.	200
--no-filtering	Disables filtering of 'Unknown' reference objects in distance queries.	False

📂 Output Structure

Results are saved in your specified --output_dir under a subfolder named after the --model_type.

results/{model_type}/
├── csv/          # Delta heatmaps, prediction rows, and cross-scale summaries
├── json/         # Consistency metrics, alignments, and validity checks per scale
├── npz/          # Raw hidden states and delta vectors for offline analysis
└── plots/        # Visualizations
    ├── all/               # Unfiltered analysis plots (PCA, bar charts, heatmaps)
    ├── both_correct/      # Strict analysis plots (only pairs where model got both right)
    └── accuracy/          # Grouped and per-category accuracy bar/line charts

🛠 Prerequisites

• torch
• transformers
• pandas, numpy, scikit-learn
• matplotlib, seaborn, tqdm
• Pillow
• Model-specific libraries: qwen_vl_utils, llava, olmo (depending on the models being tested).