Add files using upload-large-folder tool

.gitattributes

@@ -1,35 +1,6 @@
-*.7z filter=lfs diff=lfs merge=lfs -text
-*.arrow filter=lfs diff=lfs merge=lfs -text
-*.bin filter=lfs diff=lfs merge=lfs -text
-*.bz2 filter=lfs diff=lfs merge=lfs -text
-*.ckpt filter=lfs diff=lfs merge=lfs -text
-*.ftz filter=lfs diff=lfs merge=lfs -text
-*.gz filter=lfs diff=lfs merge=lfs -text
-*.h5 filter=lfs diff=lfs merge=lfs -text
-*.joblib filter=lfs diff=lfs merge=lfs -text
-*.lfs.* filter=lfs diff=lfs merge=lfs -text
-*.mlmodel filter=lfs diff=lfs merge=lfs -text
-*.model filter=lfs diff=lfs merge=lfs -text
-*.msgpack filter=lfs diff=lfs merge=lfs -text
-*.npy filter=lfs diff=lfs merge=lfs -text
-*.npz filter=lfs diff=lfs merge=lfs -text
-*.onnx filter=lfs diff=lfs merge=lfs -text
-*.ot filter=lfs diff=lfs merge=lfs -text
-*.parquet filter=lfs diff=lfs merge=lfs -text
-*.pb filter=lfs diff=lfs merge=lfs -text
-*.pickle filter=lfs diff=lfs merge=lfs -text
-*.pkl filter=lfs diff=lfs merge=lfs -text
+# Hugging Face: track large checkpoints with Git LFS
 *.pt filter=lfs diff=lfs merge=lfs -text
 *.pth filter=lfs diff=lfs merge=lfs -text
-*.rar filter=lfs diff=lfs merge=lfs -text
 *.safetensors filter=lfs diff=lfs merge=lfs -text
-saved_model/**/* filter=lfs diff=lfs merge=lfs -text
-*.tar.* filter=lfs diff=lfs merge=lfs -text
-*.tar filter=lfs diff=lfs merge=lfs -text
-*.tflite filter=lfs diff=lfs merge=lfs -text
-*.tgz filter=lfs diff=lfs merge=lfs -text
-*.wasm filter=lfs diff=lfs merge=lfs -text
-*.xz filter=lfs diff=lfs merge=lfs -text
-*.zip filter=lfs diff=lfs merge=lfs -text
-*.zst filter=lfs diff=lfs merge=lfs -text
-*tfevents* filter=lfs diff=lfs merge=lfs -text
+*.bin filter=lfs diff=lfs merge=lfs -text
+*.ckpt filter=lfs diff=lfs merge=lfs -text

README.md

@@ -0,0 +1,82 @@
+---
+license: mit
+language:
+- en
+library_name: pytorch
+tags:
+- sparse-autoencoder
+- sae
+- vision-language
+- multimodal
+- multimodal-retrieval
+- interpretability
+- mllm
+---
+
+# SAE Checkpoints for *Generative Giants, Retrieval Weaklings*
+
+[![ACL 2026](https://img.shields.io/badge/ACL-2026-b31b1b?style=flat-square)](#)
+[![arXiv](https://img.shields.io/badge/arXiv-coming%20soon-b31b1b?style=flat-square&logo=arxiv)](#)
+[![GitHub](https://img.shields.io/badge/GitHub-mllm--retrieval--analysis-1f6feb?style=flat-square&logo=github)](https://github.com/Heinz217/mllm-retrieval-analysis)
+
+Top-K sparse autoencoders (SAEs) trained on the
+[COCO-Caption](https://huggingface.co/datasets/lmms-lab/COCO-Caption) dataset,
+one checkpoint per vision / vision–language backbone. These weights accompany
+the paper:
+
+> **Generative Giants, Retrieval Weaklings: Why do Multimodal Large Language Models Fail at Multimodal Retrieval?**
+> *Accepted to ACL 2026.*
+
+The SAEs are used in the paper to dissect the representation space of MLLMs to compute the four diagnostic metrics: *energy*,
+*modality score*, *bridge score*, and *retrieval attribution score*.
+
+Code, training scripts, and analysis pipeline are released at
+[Heinz217/mllm-retrieval-analysis](https://github.com/Heinz217/mllm-retrieval-analysis).
+
+## Files
+
+| Filename                          | Backbone                  |
+|-----------------------------------|---------------------------|
+| `sae_Qwen2-VL-7B-Instruct.pt`     | Qwen2-VL-7B-Instruct      |
+| `sae_Qwen3-VL-8B-Instruct.pt`     | Qwen3-VL-8B-Instruct      |
+| `sae_paligemma-3b-mix-224.pt`     | PaliGemma2-3B-Mix-224     |
+| `sae_chameleon-7b.pt`             | Chameleon-7B              |
+| `sae_clip.pt`                     | CLIP                      |
+| `sae_siglip2.pt`                  | SigLIP2                   |
+
+A machine-readable index is provided in
+[`model_index.json`](model_index.json).
+
+## Usage
+
+Each checkpoint is a plain PyTorch state dict produced by `torch.save`. Pair it
+with the matching `TopKSAE` instance from the
+[overcomplete](https://github.com/KempnerInstitute/overcomplete) framework
+(also vendored in the [analysis repo](https://github.com/Heinz217/mllm-retrieval-analysis)):
+
+```python
+import torch
+from huggingface_hub import hf_hub_download
+from overcomplete.sae.topk_sae import TopKSAE
+
+ckpt_path = hf_hub_download(
+    repo_id="Heinz217/mllm-retrieval-analysis-sae",
+    filename="sae_Qwen2-VL-7B-Instruct.pt",
+)
+ckpt = torch.load(ckpt_path, map_location="cuda")
+
+# Initialize TopKSAE with the same input_shape / nb_concepts that were used
+# at training time, then load the trained weights.
+sae = TopKSAE(input_shape=..., nb_concepts=..., top_k=50, device="cuda").to("cuda")
+sae.load_state_dict(ckpt["model_state"])
+```
+
+For the full analysis pipeline, follow the
+instructions in the [official repository](https://github.com/Heinz217/mllm-retrieval-analysis).
+
+## License
+
+Released under the [MIT License](https://opensource.org/license/mit/), matching
+the accompanying code. The licenses of the *original* MLLM/VLM checkpoints
+(Qwen-VL, PaliGemma2, Chameleon, CLIP, SigLIP2) on which these SAEs were
+trained remain governed by their respective upstream terms.

model_index.json

@@ -0,0 +1,36 @@
+{
+  "format": "sae_torch_state_dict",
+  "description": "Sparse autoencoder checkpoints; filenames are sae_<backbone>.pt",
+  "artifacts": [
+    {
+      "filename": "sae_chameleon-7b.pt",
+      "backbone_model": "chameleon-7b",
+      "size_bytes": 3221624817
+    },
+    {
+      "filename": "sae_paligemma-3b-mix-224.pt",
+      "backbone_model": "paligemma-3b-mix-224",
+      "size_bytes": 1146872779
+    },
+    {
+      "filename": "sae_Qwen3-VL-8B-Instruct.pt",
+      "backbone_model": "Qwen3-VL-8B-Instruct",
+      "size_bytes": 3221624779
+    },
+    {
+      "filename": "sae_Qwen2-VL-7B-Instruct.pt",
+      "backbone_model": "Qwen2-VL-7B-Instruct",
+      "size_bytes": 2818971659
+    },
+    {
+      "filename": "sae_siglip2.pt",
+      "backbone_model": "siglip2",
+      "size_bytes": 129520023
+    },
+    {
+      "filename": "sae_clip.pt",
+      "backbone_model": "clip",
+      "size_bytes": 129520049
+    }
+  ]
+}

sae_Qwen2-VL-7B-Instruct.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:169947f23b84fca2279f34caea134ba960f9457b3aaead37b08cdf5b2f8637c3
+size 2818971659

sae_Qwen3-VL-8B-Instruct.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b8d05ab1395d082e042fc35486bc844fabd779e0dd0bc686f190578c593a6643
+size 3221624779

sae_chameleon-7b.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:94fc469cf22673419ff60b44c944636a4a57d34f7f9bbb713cefc6a31f77af04
+size 3221624817

sae_clip.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e7f853b1b7bcf2b77b410837a4dbd7e779ea21c18de8e329ccd5eb9d5e7550b0
+size 129520049

sae_paligemma-3b-mix-224.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8514c3c7a623d42bc9ad23dd949bb3615baab460f665a452625ab63e07cfc458
+size 1146872779

sae_siglip2.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c14b14d451d2c97c770c80017fd1f5adaca801204d7bcbf0497b0c64ed2d474b
+size 129520023