init

.gitattributes

@@ -0,0 +1,60 @@
+*.7z filter=lfs diff=lfs merge=lfs -text
+*.arrow filter=lfs diff=lfs merge=lfs -text
+*.avro 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
+*.lz4 filter=lfs diff=lfs merge=lfs -text
+*.mds 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
+*.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
+# Audio files - uncompressed
+*.pcm filter=lfs diff=lfs merge=lfs -text
+*.sam filter=lfs diff=lfs merge=lfs -text
+*.raw filter=lfs diff=lfs merge=lfs -text
+# Audio files - compressed
+*.aac filter=lfs diff=lfs merge=lfs -text
+*.flac filter=lfs diff=lfs merge=lfs -text
+*.mp3 filter=lfs diff=lfs merge=lfs -text
+*.ogg filter=lfs diff=lfs merge=lfs -text
+*.wav filter=lfs diff=lfs merge=lfs -text
+# Image files - uncompressed
+*.bmp filter=lfs diff=lfs merge=lfs -text
+*.gif filter=lfs diff=lfs merge=lfs -text
+*.png filter=lfs diff=lfs merge=lfs -text
+*.tiff filter=lfs diff=lfs merge=lfs -text
+# Image files - compressed
+*.jpg filter=lfs diff=lfs merge=lfs -text
+*.jpeg filter=lfs diff=lfs merge=lfs -text
+*.webp filter=lfs diff=lfs merge=lfs -text
+# Video files - compressed
+*.mp4 filter=lfs diff=lfs merge=lfs -text
+*.webm filter=lfs diff=lfs merge=lfs -text

README.md

@@ -0,0 +1,75 @@
+---
+pretty_name: SMILE artifacts
+language:
+- en
+tags:
+- in-context-learning
+- vision-language-models
+- hard-negatives
+task_categories:
+- visual-question-answering
+- image-classification
+viewer: false
+---
+
+# SMILE: CLIP embeddings, mined hard-negative pools and predictions
+
+Artifacts for **The Value of Being Wrong: Self-Mined Visual In-Context Learning from Errors** (NeurIPS 2026 Spotlight).
+Code: [github.com/Tree-Shu-Zhao/SMILE](https://github.com/Tree-Shu-Zhao/SMILE).
+
+## Contents
+
+```
+clip_embeddings/
+  <data>_<split>_image.pt            # CLIP ViT-g-14 (laion2b_s12b_b42k) image embeddings, L2-normalized
+  <data>_<split>_ids.pt              # the sample id of each row
+hard_negatives/<data>/<org_model>/
+  hard_negatives.jsonl               # the pool: training samples the model gets wrong, with its voted prediction
+  hard_negatives_predictions.jsonl   # the voted prediction for every training sample
+  hard_negatives_stats.json
+  hard_negatives_params.json
+predictions/<data>/<org_model>/
+  vicl_<k>_<negative_k>.jsonl        # test predictions, with retrieved_ids and contrastive_ids
+  vicl_<k>_<negative_k>_params.json
+```
+
+- `<data>`: `cub`, `flowers`, `okvqa`, `vizwiz`. `<split>`: `train`, `test` (for VizWiz, `test` is its val split).
+- `<org_model>`: `Qwen_Qwen2.5-VL-7B-Instruct`, `Qwen_Qwen2.5-VL-3B-Instruct`, `google_gemma-3-12b-it`,
+  `google_gemma-3-4b-it`.
+- Pool records: `id`, `question`, `imgpath`, `gt_answer`, `voted_pred`, `vote_count`, `total_votes`, `token_f1`,
+  `is_correct`.
+- Prediction records: `id`, `question`, `imgpath`, `pred_answer`, `gt_answer`, `k`, `retrieved_ids`,
+  `contrastive_ids`. The VizWiz records hold `id`, `pred_answer`, `query_prediction`, `retrieved_ids` and
+  `contrastive_ids`.
+- The predictions are one inference run per model and dataset. VizWiz predictions are stored before the
+  answerability vote, which the scorer applies.
+- `imgpath` is relative to the root of the code repository (`data/<dataset>/...`).
+- The records contain questions and answers from CUB-200-2011, Oxford 102 Flowers, OK-VQA and VizWiz, which remain
+  under the terms of those datasets. No images are included.
+
+## Usage
+
+From the root of the code repository:
+
+```bash
+uvx --from "huggingface_hub>=1.0" hf download TreezzZ/SMILE-data --repo-type dataset --local-dir . \
+  --include "clip_embeddings/*" --include "hard_negatives/*" --include "predictions/*"
+```
+
+The files land where the code reads them. To score the released predictions:
+
+```bash
+bash scripts/cub/evaluate.sh predictions/cub/Qwen_Qwen2.5-VL-7B-Instruct/vicl_8_4.jsonl
+bash scripts/vizwiz/evaluate.sh predictions/vizwiz/Qwen_Qwen2.5-VL-7B-Instruct/vicl_12_8.jsonl
+```
+
+## Citation
+
+```bibtex
+@inproceedings{zhao2026smile,
+  title={The Value of Being Wrong: Self-Mined Visual In-Context Learning from Errors},
+  author={Zhao, Shu and Yu, Tan},
+  booktitle={Advances in Neural Information Processing Systems},
+  year={2026}
+}
+```

clip_embeddings/cub_test_ids.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:aedae8459f857541f2eba9ffe215beed1b0622f84eac691c4096e0243397a611
+size 81588

clip_embeddings/cub_test_image.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:02e93b2a80075af7db2566f1f865268de9402cd1975001644bd54269539f9b41
+size 23733439

clip_embeddings/cub_train_ids.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e1aef73a1a140e174056ed3af4f090f345bd6a7b07b51fcaf9e125f5c946bec9
+size 84344

clip_embeddings/cub_train_image.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e28e3327597c58fb921d6e1e77c5757be0c57314fafa0b423e7d9523c224d3f6
+size 24552644

clip_embeddings/flowers_test_ids.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:87cf58db9cda3cf7bb56897d38dadeec82f6bfae2969bc856e6c2aa2d653176b
+size 85252

clip_embeddings/flowers_test_image.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c43ebab7d4f3106ea586e23a23108fdb1aaf28867743844c5f71c47a48d78e59
+size 25187539

clip_embeddings/flowers_train_ids.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e9bd0e71f6c7b315b6796bd668744be5dfacd5f87177505b201efb5e81519ee0
+size 28552

clip_embeddings/flowers_train_image.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ff90785d538776332177403250c7497caa0fe251621583eeeaa75c58ee41a40f
+size 8357080

clip_embeddings/okvqa_test_ids.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:32330bbb692193ceaaf678cf8805e8e0d6f00d9aae309d9ae0ed325ae889ece4
+size 84988

clip_embeddings/okvqa_test_image.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b91c749b4bc997d79ef3d69e7f1c60ae30aa3da2d90c7be4649a11aed7a3e5c3
+size 20669641

clip_embeddings/okvqa_train_ids.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5e35d48eac70cc11ea3e86339b0cac5d567d2eaf6d3865c4135f96fff20f6e6c
+size 151488

clip_embeddings/okvqa_train_image.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f62658059e334165fc015108a4328d03cc7f29957740f6cba0f8f64cf9ccb275
+size 36902094

clip_embeddings/vizwiz_test_ids.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5e83f80e06c8bc65953fdd3149a6fb3d77579761814cd55f4229b10091a1e039
+size 142656

clip_embeddings/vizwiz_test_image.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:491032917d9fedd38ca1c4850347211fb5e8fb9efd60c6068ad9f5678515f9a1
+size 17691854

clip_embeddings/vizwiz_train_ids.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0fb29523063f58d1d942cb7409ec206832f04cdc4e15af0194779ee538cfd43b
+size 718468

clip_embeddings/vizwiz_train_image.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c191e471598befe9b0a405120b85097ddce3fbc726a635373a8401d0860639a8
+size 84063443

hard_negatives/cub/Qwen_Qwen2.5-VL-3B-Instruct/hard_negatives_params.json

@@ -0,0 +1,5 @@
+{
+  "model": "Qwen/Qwen2.5-VL-3B-Instruct",
+  "data": "cub",
+  "num_votes": 3
+}

hard_negatives/cub/Qwen_Qwen2.5-VL-3B-Instruct/hard_negatives_stats.json

@@ -0,0 +1,22 @@
+{
+  "total_train_samples": 5994,
+  "total_processed": 5994,
+  "total_correct": 814,
+  "total_incorrect": 5180,
+  "hard_negatives_count": 5180,
+  "num_votes": 3,
+  "token_f1": {
+    "mean": 0.2508,
+    "median": 0.0,
+    "min": 0.0,
+    "max": 1.0
+  },
+  "vote_count": {
+    "mean": 1.7681,
+    "distribution": {
+      "2": 2270,
+      "3": 1167,
+      "1": 2557
+    }
+  }
+}

hard_negatives/cub/Qwen_Qwen2.5-VL-7B-Instruct/hard_negatives_params.json

@@ -0,0 +1,5 @@
+{
+  "model": "Qwen/Qwen2.5-VL-7B-Instruct",
+  "data": "cub",
+  "num_votes": 3
+}

hard_negatives/cub/Qwen_Qwen2.5-VL-7B-Instruct/hard_negatives_stats.json

@@ -0,0 +1,22 @@
+{
+  "total_train_samples": 5994,
+  "total_processed": 5994,
+  "total_correct": 1100,
+  "total_incorrect": 4894,
+  "hard_negatives_count": 4894,
+  "num_votes": 3,
+  "token_f1": {
+    "mean": 0.3325,
+    "median": 0.0,
+    "min": 0.0,
+    "max": 1.0
+  },
+  "vote_count": {
+    "mean": 2.1975,
+    "distribution": {
+      "3": 2371,
+      "2": 2436,
+      "1": 1187
+    }
+  }
+}

hard_negatives/cub/google_gemma-3-12b-it/hard_negatives_params.json

@@ -0,0 +1,5 @@
+{
+  "model": "google/gemma-3-12b-it",
+  "data": "cub",
+  "num_votes": 3
+}

hard_negatives/cub/google_gemma-3-12b-it/hard_negatives_stats.json

@@ -0,0 +1,22 @@
+{
+  "total_train_samples": 5994,
+  "total_processed": 5994,
+  "total_correct": 1901,
+  "total_incorrect": 4093,
+  "hard_negatives_count": 4093,
+  "num_votes": 3,
+  "token_f1": {
+    "mean": 0.4882,
+    "median": 0.5,
+    "min": 0.0,
+    "max": 1.0
+  },
+  "vote_count": {
+    "mean": 2.8126,
+    "distribution": {
+      "3": 4958,
+      "2": 949,
+      "1": 87
+    }
+  }
+}

hard_negatives/cub/google_gemma-3-4b-it/hard_negatives_params.json

@@ -0,0 +1,5 @@
+{
+  "model": "google/gemma-3-4b-it",
+  "data": "cub",
+  "num_votes": 3
+}

hard_negatives/cub/google_gemma-3-4b-it/hard_negatives_stats.json

@@ -0,0 +1,22 @@
+{
+  "total_train_samples": 5994,
+  "total_processed": 5994,
+  "total_correct": 850,
+  "total_incorrect": 5144,
+  "hard_negatives_count": 5144,
+  "num_votes": 3,
+  "token_f1": {
+    "mean": 0.2778,
+    "median": 0.0,
+    "min": 0.0,
+    "max": 1.0
+  },
+  "vote_count": {
+    "mean": 2.8395,
+    "distribution": {
+      "3": 5096,
+      "2": 834,
+      "1": 64
+    }
+  }
+}

hard_negatives/flowers/Qwen_Qwen2.5-VL-3B-Instruct/hard_negatives_params.json

@@ -0,0 +1,5 @@
+{
+  "model": "Qwen/Qwen2.5-VL-3B-Instruct",
+  "data": "flowers",
+  "num_votes": 3
+}

hard_negatives/flowers/Qwen_Qwen2.5-VL-3B-Instruct/hard_negatives_stats.json

@@ -0,0 +1,22 @@
+{
+  "total_train_samples": 2040,
+  "total_processed": 2040,
+  "total_correct": 387,
+  "total_incorrect": 1653,
+  "hard_negatives_count": 1653,
+  "num_votes": 3,
+  "token_f1": {
+    "mean": 0.25,
+    "median": 0.0,
+    "min": 0.0,
+    "max": 1.0
+  },
+  "vote_count": {
+    "mean": 1.7583,
+    "distribution": {
+      "3": 381,
+      "1": 874,
+      "2": 785
+    }
+  }
+}

hard_negatives/flowers/Qwen_Qwen2.5-VL-7B-Instruct/hard_negatives_params.json

@@ -0,0 +1,5 @@
+{
+  "model": "Qwen/Qwen2.5-VL-7B-Instruct",
+  "data": "flowers",
+  "num_votes": 3
+}

hard_negatives/flowers/Qwen_Qwen2.5-VL-7B-Instruct/hard_negatives_stats.json

@@ -0,0 +1,22 @@
+{
+  "total_train_samples": 2040,
+  "total_processed": 2040,
+  "total_correct": 764,
+  "total_incorrect": 1276,
+  "hard_negatives_count": 1276,
+  "num_votes": 3,
+  "token_f1": {
+    "mean": 0.4411,
+    "median": 0.0,
+    "min": 0.0,
+    "max": 1.0
+  },
+  "vote_count": {
+    "mean": 2.2877,
+    "distribution": {
+      "1": 349,
+      "3": 936,
+      "2": 755
+    }
+  }
+}

hard_negatives/flowers/google_gemma-3-12b-it/hard_negatives_params.json

@@ -0,0 +1,5 @@
+{
+  "model": "google/gemma-3-12b-it",
+  "data": "flowers",
+  "num_votes": 3
+}

hard_negatives/flowers/google_gemma-3-12b-it/hard_negatives_stats.json

@@ -0,0 +1,22 @@
+{
+  "total_train_samples": 2040,
+  "total_processed": 2040,
+  "total_correct": 1292,
+  "total_incorrect": 748,
+  "hard_negatives_count": 748,
+  "num_votes": 3,
+  "token_f1": {
+    "mean": 0.6603,
+    "median": 1.0,
+    "min": 0.0,
+    "max": 1.0
+  },
+  "vote_count": {
+    "mean": 2.9279,
+    "distribution": {
+      "3": 1910,
+      "2": 113,
+      "1": 17
+    }
+  }
+}

hard_negatives/flowers/google_gemma-3-4b-it/hard_negatives_params.json

@@ -0,0 +1,5 @@
+{
+  "model": "google/gemma-3-4b-it",
+  "data": "flowers",
+  "num_votes": 3
+}

hard_negatives/flowers/google_gemma-3-4b-it/hard_negatives_stats.json

@@ -0,0 +1,22 @@
+{
+  "total_train_samples": 2040,
+  "total_processed": 2040,
+  "total_correct": 719,
+  "total_incorrect": 1321,
+  "hard_negatives_count": 1321,
+  "num_votes": 3,
+  "token_f1": {
+    "mean": 0.3955,
+    "median": 0.0,
+    "min": 0.0,
+    "max": 1.0
+  },
+  "vote_count": {
+    "mean": 2.9216,
+    "distribution": {
+      "3": 1886,
+      "1": 6,
+      "2": 148
+    }
+  }
+}