Datasets:
Expose sampled negative item metadata and expand dataset documentation
Browse files- README.md +139 -69
- manifest.json +12 -2
- reports/negative_metadata_coverage.json +34 -0
- tables/negative_items/test.parquet +3 -0
README.md
CHANGED
|
@@ -23,6 +23,10 @@ configs:
|
|
| 23 |
data_files:
|
| 24 |
- split: test
|
| 25 |
path: tables/products/*.parquet
|
|
|
|
|
|
|
|
|
|
|
|
|
| 26 |
- config_name: users
|
| 27 |
data_files:
|
| 28 |
- split: test
|
|
@@ -37,58 +41,108 @@ configs:
|
|
| 37 |
path: tables/evaluation_candidates/*.parquet
|
| 38 |
---
|
| 39 |
|
| 40 |
-
# MemoryVCD
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 41 |
|
| 42 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
| 43 |
|
| 44 |
-
|
| 45 |
|
| 46 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 47 |
|
| 48 |
-
|
| 49 |
|
| 50 |
-
|
| 51 |
|
| 52 |
-
| Subset | Rows |
|
| 53 |
|---|---:|---|
|
| 54 |
-
| [interactions](https://huggingface.co/datasets/WZDavid/MemoryVCD/viewer/interactions/test) | 535,698 |
|
| 55 |
-
| [products](https://huggingface.co/datasets/WZDavid/MemoryVCD/viewer/products/test) | 481,333 |
|
| 56 |
-
| [
|
| 57 |
-
| [
|
| 58 |
-
| [
|
|
|
|
| 59 |
|
| 60 |
-
|
| 61 |
|
| 62 |
-
##
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 63 |
|
| 64 |
```python
|
| 65 |
from datasets import load_dataset
|
| 66 |
|
| 67 |
repo = "WZDavid/MemoryVCD"
|
| 68 |
reviews = load_dataset(repo, "interactions", split="test")
|
| 69 |
-
|
| 70 |
-
|
| 71 |
-
print(reviews[0])
|
| 72 |
-
```
|
| 73 |
|
| 74 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
| 75 |
|
| 76 |
-
|
| 77 |
|
| 78 |
```python
|
| 79 |
-
|
| 80 |
-
|
| 81 |
-
|
| 82 |
-
|
| 83 |
-
|
| 84 |
-
users = [json.loads(line) for line in f if line.strip()]
|
| 85 |
```
|
| 86 |
|
| 87 |
-
|
| 88 |
-
|
| 89 |
-
## Corpus distribution
|
| 90 |
|
| 91 |
-
The first four domains are evaluation targets
|
| 92 |
|
| 93 |
| Domain | Users | Interactions | History products |
|
| 94 |
|---|---:|---:|---:|
|
|
@@ -104,14 +158,11 @@ The first four domains are evaluation targets; the remaining eight provide addit
|
|
| 104 |
| Movies_and_TV | 278 | 9,395 | 8,433 |
|
| 105 |
| Office_Products | 323 | 28,325 | 14,032 |
|
| 106 |
| Sports_and_Outdoors | 323 | 26,847 | 15,194 |
|
|
|
|
| 107 |
|
| 108 |
-
##
|
| 109 |
-
|
| 110 |
-
All **481,333 required product/domain records** have metadata. This includes the random ranking candidates and all saved time-cutoff hard candidates, beyond the 276,418 history products.
|
| 111 |
|
| 112 |
-
|
| 113 |
-
|
| 114 |
-
| Domain | Required products | Image URLs | Cached images |
|
| 115 |
|---|---:|---:|---:|
|
| 116 |
| Beauty_and_Personal_Care | 82,891 | 82,890 | 49,225 |
|
| 117 |
| Books | 100,625 | 78,730 | 49,539 |
|
|
@@ -125,38 +176,57 @@ There are **456,761 product image URLs** and **243,394 valid cached product imag
|
|
| 125 |
| Movies_and_TV | 8,433 | 5,840 | 244 |
|
| 126 |
| Office_Products | 14,032 | 13,977 | 3,936 |
|
| 127 |
| Sports_and_Outdoors | 15,194 | 15,194 | 5,485 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 128 |
|
| 129 |
-
|
| 130 |
-
|
| 131 |
-
Product images differ from user-uploaded **review images**. The primary-domain histories preserve their review-image fields; the eight added domains' extraction did not retain these fields. A missing field there must not be interpreted as proof that the original review had no image. `review_images_field_present` makes this distinction explicit. Review image bytes are not bundled.
|
| 132 |
-
|
| 133 |
-
## Fields and ordering
|
| 134 |
-
|
| 135 |
-
- `interaction_id`: stable identifier formed from source user index, domain, and original position within that domain.
|
| 136 |
-
- `interaction_index`: zero-based index after a stable timestamp sort within one user/domain; used by evaluation windows.
|
| 137 |
-
- `source_user_index` and `source_interaction_index`: preserve source ordering and allow reconstruction.
|
| 138 |
-
- `timestamp`: original Unix milliseconds; `date` is its UTC ISO representation.
|
| 139 |
-
- `review_title` and `review_text`: complete review text. `rating` preserves the original numeric score.
|
| 140 |
-
- `image_url`: selected MAIN product image (large, then hi_res, then thumb; fallback to another variant).
|
| 141 |
-
- `review_images_json`: JSON representation of original review images; `null` if absent.
|
| 142 |
-
- Product metadata mirrors the evaluator's slim representation: title limited to 200 characters and description to 400. No `price` field is promised in the new product tables. Per-domain slim JSONL files are in `metadata/`.
|
| 143 |
-
|
| 144 |
-
## Evaluation protocol and actual candidate coverage
|
| 145 |
-
|
| 146 |
-
All 323 users have windows for the four primary target domains. Time cutoffs are 0, .25, .5, .75 and 1. A three-interaction window moves along the timestamp-sorted target history; preceding interactions are the initial memory. Interior centers use Python `round(c*n)`, with indices `[center-1, center+2)`.
|
| 147 |
-
|
| 148 |
-
Domain windows hold the target's final three interactions fixed. D0 has no initial memory. D1–D3 add non-target primary domains in the order Personal Care, Books, Electronics, Home & Kitchen, skipping the target. D4 adds the target; D5 adds the eight additional domains; Dtx uses only the additional eight. Memory must precede the earliest test timestamp **strictly**, and source-domain segments remain in their specified order. For Image Prediction, earlier test interactions are revealed before later predictions; other tasks retain the same initial memory throughout the window.
|
| 149 |
-
|
| 150 |
-
**Saved candidates have different coverage from windows.** The `data/v8e/...` source files cover the original first 100 users at each cutoff. The base `data/cross_domain.jsonl` ranking candidates and `hard_negatives_*_v3.jsonl` choice candidates are separately identified by `source_file`. Missing saved candidates are not fabricated for the other users. Filter by `task`, `source_file`, and `cutoff` before evaluation; do not mix candidate versions.
|
| 151 |
-
|
| 152 |
-
The stored hard-negative records contain **3 negatives**, although a paper description says 4. This release preserves the actual files rather than silently adding a fourth candidate. Ranking pools retain the saved negatives. `evaluation/source_candidates/` also contains unchanged hard-negative JSONL files. The six tasks reuse these histories/windows; they are not six independent corpora.
|
| 153 |
-
|
| 154 |
-
## Integrity and limitations
|
| 155 |
-
|
| 156 |
-
The original histories contain **10,404 excess records sharing a user/domain/product/timestamp key**. They are retained to preserve the original evaluation timelines; this is not a claim that all 535,698 rows are unique events. Stable `interaction_id` values distinguish the retained rows. Users were selected for substantial overlap/activity across the four primary domains; they are not a representative random sample of Amazon users. Ratings are strongly skewed toward 4–5 stars.
|
| 157 |
-
|
| 158 |
-
See `reports/` for corpus, metadata, image and release validation results. `manifest.json` lists SHA-256 hashes and sizes for release files. Image caches were checked for valid decoding before export. No external image downloads were performed during packaging.
|
| 159 |
-
|
| 160 |
-
## Source and attribution
|
| 161 |
|
| 162 |
-
|
|
|
|
| 23 |
data_files:
|
| 24 |
- split: test
|
| 25 |
path: tables/products/*.parquet
|
| 26 |
+
- config_name: negative_items
|
| 27 |
+
data_files:
|
| 28 |
+
- split: test
|
| 29 |
+
path: tables/negative_items/*.parquet
|
| 30 |
- config_name: users
|
| 31 |
data_files:
|
| 32 |
- split: test
|
|
|
|
| 41 |
path: tables/evaluation_candidates/*.parquet
|
| 42 |
---
|
| 43 |
|
| 44 |
+
# 🧠 MemoryVCD · URL Edition
|
| 45 |
+
|
| 46 |
+
**323 users · 12 domains · 535,698 interactions**
|
| 47 |
+
|
| 48 |
+
[📄 Paper](https://aclanthology.org/2026.findings-emnlp.1804/) · [📥 PDF](https://aclanthology.org/2026.findings-emnlp.1804.pdf) · [💻 MemoryCD Code](https://github.com/AgentMemoryWorld/MemoryCD) · [🔗 URL Edition](https://huggingface.co/datasets/WZDavid/MemoryVCD) · [🖼️ Full Edition](https://huggingface.co/datasets/WZDavid/MemoryVCD-Full)
|
| 49 |
+
|
| 50 |
+
Real user histories for evaluating long-term, cross-domain personalization. MemoryVCD extends the MemoryCD data release with product image references, complete metadata for the saved candidate pools, and explicit evaluation inputs. Full adds the available cached product images.
|
| 51 |
+
|
| 52 |
+
## 📄 Paper and release relationship
|
| 53 |
+
|
| 54 |
+
**MemoryCD: Benchmarking Long-Context User Memory of LLM Agents for Lifelong Cross-Domain Personalization** — Zhang et al., *Findings of EMNLP 2026*.
|
| 55 |
+
|
| 56 |
+
The paper studies memory grounded in real Amazon review histories, with four tasks: rating prediction, item ranking, review summarization, and review generation. It evaluates single-domain and cross-domain memory. See the [paper](https://aclanthology.org/2026.findings-emnlp.1804/) for methods and results.
|
| 57 |
+
|
| 58 |
+
This repository is the **MemoryVCD extended release**, not a claim that every added visual feature or candidate file was evaluated in the MemoryCD paper. Statistics below describe the released files.
|
| 59 |
|
| 60 |
+
| Repository | Purpose | Product images |
|
| 61 |
+
|---|---|---|
|
| 62 |
+
| [MemoryCD](https://huggingface.co/datasets/WZDavid/MemoryCD) | Original release; unchanged | Original item metadata |
|
| 63 |
+
| [MemoryVCD](https://huggingface.co/datasets/WZDavid/MemoryVCD) | Extended histories, candidate metadata and evaluation inputs | URL strings |
|
| 64 |
+
| [MemoryVCD-Full](https://huggingface.co/datasets/WZDavid/MemoryVCD-Full) | Same data plus existing image cache | 243,394 embedded image records |
|
| 65 |
|
| 66 |
+
## 📊 At a glance
|
| 67 |
|
| 68 |
+
| Statistic | Count |
|
| 69 |
+
|---|---:|
|
| 70 |
+
| Users | 323 |
|
| 71 |
+
| Domains | 12: 4 evaluation targets + 8 memory domains |
|
| 72 |
+
| Interaction records | 535,698 |
|
| 73 |
+
| History product/domain pairs | 276,418 |
|
| 74 |
+
| All product/domain metadata records | 481,333 |
|
| 75 |
+
| Unique sampled negative product/domain pairs | 211,364 |
|
| 76 |
+
| Negative items outside the history product set | 204,915 |
|
| 77 |
+
| Missing metadata for saved negatives | **0** |
|
| 78 |
+
| Product image URLs | 456,761 |
|
| 79 |
+
| Valid cached image records (Full) | 243,394 |
|
| 80 |
|
| 81 |
+
Counts use `(domain, parent_asin)` as the item key. A negative for one user can be a historical item for another; history and negative counts therefore overlap.
|
| 82 |
|
| 83 |
+
## 🗂️ Browse the data
|
| 84 |
|
| 85 |
+
| Subset | Rows | Contents |
|
| 86 |
|---|---:|---|
|
| 87 |
+
| [interactions](https://huggingface.co/datasets/WZDavid/MemoryVCD/viewer/interactions/test) | 535,698 | One review per row |
|
| 88 |
+
| [products](https://huggingface.co/datasets/WZDavid/MemoryVCD/viewer/products/test) | 481,333 | History and candidate item metadata |
|
| 89 |
+
| [negative_items](https://huggingface.co/datasets/WZDavid/MemoryVCD/viewer/negative_items/test) | 211,364 | Metadata for all saved sampled negatives |
|
| 90 |
+
| [users](https://huggingface.co/datasets/WZDavid/MemoryVCD/viewer/users/test) | 323 | User index and activity counts |
|
| 91 |
+
| [evaluation_windows](https://huggingface.co/datasets/WZDavid/MemoryVCD/viewer/evaluation_windows/test) | 15,504 | Memory boundaries and test targets |
|
| 92 |
+
| [evaluation_candidates](https://huggingface.co/datasets/WZDavid/MemoryVCD/viewer/evaluation_candidates/test) | 19,751 | Positive and negative item IDs with source provenance |
|
| 93 |
|
| 94 |
+
All subsets use `test` to label the evaluation corpus. This is **not** a random train/test split: `evaluation_windows` defines memory and test targets within each history.
|
| 95 |
|
| 96 |
+
## 🎯 Negative items and sampling
|
| 97 |
+
|
| 98 |
+
`products` includes both historical items and all items required by the saved ranking/choice candidates. `negative_items` is a convenient filtered view of those same metadata records; it does not add new interactions or change the candidate pools.
|
| 99 |
+
|
| 100 |
+
| Domain | Unique negative items | Outside history | Missing metadata |
|
| 101 |
+
|---|---:|---:|---:|
|
| 102 |
+
| Beauty_and_Personal_Care | 51,966 | 49,416 | 0 |
|
| 103 |
+
| Books | 53,433 | 52,571 | 0 |
|
| 104 |
+
| Electronics | 52,641 | 51,075 | 0 |
|
| 105 |
+
| Home_and_Kitchen | 53,324 | 51,853 | 0 |
|
| 106 |
+
| **Total** | **211,364** | **204,915** | **0** |
|
| 107 |
+
|
| 108 |
+
| Candidate task | Saved rows | Negatives per row | Source |
|
| 109 |
+
|---|---:|---:|---|
|
| 110 |
+
| Item ranking | 9,876 | 19 | Saved random candidate pools |
|
| 111 |
+
| Item preference choice | 9,875 | 3 | Saved hard-negative pools |
|
| 112 |
+
|
| 113 |
+
**No resampling was performed.** Candidate order and IDs are preserved. The saved hard-negative files contain 3 negatives; any description expecting 4 does not match these files. Base and time-cutoff sources are distinct: filter `task`, `source_file`, and `cutoff` before evaluation. Time-axis source files cover the original first 100 users, while evaluation windows cover all 323 users. Missing candidate rows are not fabricated.
|
| 114 |
+
|
| 115 |
+
Join `evaluation_candidates.neg_asins` to `negative_items.parent_asin` **within the same domain**. `candidate_reference_count` counts occurrences across all saved candidate rows; `candidate_tasks` lists the tasks using each item. The coverage audit is in `reports/negative_metadata_coverage.json`.
|
| 116 |
+
|
| 117 |
+
## 🚀 Load and join
|
| 118 |
|
| 119 |
```python
|
| 120 |
from datasets import load_dataset
|
| 121 |
|
| 122 |
repo = "WZDavid/MemoryVCD"
|
| 123 |
reviews = load_dataset(repo, "interactions", split="test")
|
| 124 |
+
candidates = load_dataset(repo, "evaluation_candidates", split="test")
|
| 125 |
+
negatives = load_dataset(repo, "negative_items", split="test")
|
|
|
|
|
|
|
| 126 |
|
| 127 |
+
meta = {(r["domain"], r["parent_asin"]): r for r in negatives}
|
| 128 |
+
example = candidates[0] # Select your intended task/source/cutoff for evaluation.
|
| 129 |
+
negative_metadata = [meta[(example["domain"], a)] for a in example["neg_asins"]]
|
| 130 |
+
print([(r["parent_asin"], r["title"]) for r in negative_metadata])
|
| 131 |
+
```
|
| 132 |
|
| 133 |
+
For positive item metadata, use the `products` subset. To read cached images:
|
| 134 |
|
| 135 |
```python
|
| 136 |
+
images = load_dataset("WZDavid/MemoryVCD-Full", "product_images",
|
| 137 |
+
split="test", streaming=True)
|
| 138 |
+
example_image = next(iter(images))
|
| 139 |
+
print(example_image["domain"], example_image["parent_asin"])
|
| 140 |
+
image = example_image["image"] # PIL image; cached bytes are embedded in Parquet.
|
|
|
|
| 141 |
```
|
| 142 |
|
| 143 |
+
## 🌍 Domain statistics
|
|
|
|
|
|
|
| 144 |
|
| 145 |
+
The first four domains are evaluation targets (314,116 interactions). The remaining eight provide additional memory (221,582 interactions). Users overlap across domains.
|
| 146 |
|
| 147 |
| Domain | Users | Interactions | History products |
|
| 148 |
|---|---:|---:|---:|
|
|
|
|
| 158 |
| Movies_and_TV | 278 | 9,395 | 8,433 |
|
| 159 |
| Office_Products | 323 | 28,325 | 14,032 |
|
| 160 |
| Sports_and_Outdoors | 323 | 26,847 | 15,194 |
|
| 161 |
+
| **Total (unique users)** | **323** | **535,698** | **276,418** |
|
| 162 |
|
| 163 |
+
## 🖼️ Images and metadata
|
|
|
|
|
|
|
| 164 |
|
| 165 |
+
| Domain | Required metadata | Image URLs | Cached image records |
|
|
|
|
|
|
|
| 166 |
|---|---:|---:|---:|
|
| 167 |
| Beauty_and_Personal_Care | 82,891 | 82,890 | 49,225 |
|
| 168 |
| Books | 100,625 | 78,730 | 49,539 |
|
|
|
|
| 176 |
| Movies_and_TV | 8,433 | 5,840 | 244 |
|
| 177 |
| Office_Products | 14,032 | 13,977 | 3,936 |
|
| 178 |
| Sports_and_Outdoors | 15,194 | 15,194 | 5,485 |
|
| 179 |
+
| **Total** | **481,333** | **456,761** | **243,394** |
|
| 180 |
+
|
| 181 |
+
Full includes all matched valid images in the release cache, **not 100% image coverage**: 50.57% of required metadata rows and 30.23% of history products have cached images. One invalid cached image was excluded. External URLs may expire. Product photos are separate from user-uploaded review photos; review image bytes are not bundled.
|
| 182 |
+
|
| 183 |
+
## 🧾 Key fields and files
|
| 184 |
+
|
| 185 |
+
| Table / file | Important fields or contents |
|
| 186 |
+
|---|---|
|
| 187 |
+
| `interactions` | `user_id`, `domain`, `parent_asin`, `rating`, `timestamp`, review title/text, `image_url` |
|
| 188 |
+
| `products` / `negative_items` | `title`, `description`, category, store, average rating, image URL; `in_user_history`, cache availability |
|
| 189 |
+
| `evaluation_candidates` | `task`, `source_file`, `cutoff`, `pos_asin`, `neg_asins` |
|
| 190 |
+
| `evaluation_windows` | Target indices, memory prefix/boundary, source domains |
|
| 191 |
+
| `product_images` (Full) | `(domain, parent_asin)`, embedded `image`, source URL and image hash |
|
| 192 |
+
| `users/users_12domains.jsonl.gz` | Original nested histories, one user per line |
|
| 193 |
+
| `metadata/*.jsonl.gz` | Per-domain metadata including required candidate items |
|
| 194 |
+
| `evaluation/source_candidates/` | Original saved hard-negative files |
|
| 195 |
+
| `tools/release_dataset_helpers.py` | Load nested histories and reconstruct evaluation inputs |
|
| 196 |
+
| `reports/`, `manifest.json` | Coverage checks, validation, file sizes and SHA-256 hashes |
|
| 197 |
+
|
| 198 |
+
Metadata uses the evaluator's slim fields: title up to 200 characters and description up to 400; price is not included. `image_shard` paths refer to the Full repository. Join images with `(domain, parent_asin)`. The `users` subset is an index; use the compressed histories for the complete nested records.
|
| 199 |
+
|
| 200 |
+
## ⚙️ Evaluation and limitations
|
| 201 |
+
|
| 202 |
+
| Setting | Released definition |
|
| 203 |
+
|---|---|
|
| 204 |
+
| Time axis | Five cutoffs: 0, .25, .5, .75, 1; a three-interaction target window |
|
| 205 |
+
| Domain axis | D0–D5 and Dtx; final three target-domain interactions |
|
| 206 |
+
| Chronology | Memory precedes the target boundary; see helper functions for exact reconstruction |
|
| 207 |
+
| Ordering | `source_user_index`, `source_interaction_index` and stable interaction IDs preserve provenance |
|
| 208 |
+
| Repeated keys | 10,404 excess records share user/domain/item/timestamp; retained to preserve timelines |
|
| 209 |
+
| Selection bias | Users have substantial cross-domain activity; not a representative random sample |
|
| 210 |
+
| Review images | Extra-eight-domain extraction did not retain review-image fields; absence is not proof of no original image |
|
| 211 |
+
|
| 212 |
+
The same histories/windows support multiple tasks; task counts must not be summed as independent users. Exact source ordering and candidate versions matter for reproducibility.
|
| 213 |
+
|
| 214 |
+
## 📚 Citation and source
|
| 215 |
+
|
| 216 |
+
Please cite the foundational MemoryCD paper and identify the MemoryVCD edition/revision used:
|
| 217 |
+
|
| 218 |
+
```bibtex
|
| 219 |
+
@inproceedings{zhang-etal-2026-memorycd,
|
| 220 |
+
title = "{M}emory{CD}: Benchmarking Long-Context User Memory of {LLM} Agents for Lifelong Cross-Domain Personalization",
|
| 221 |
+
author = "Zhang, Weizhi and Wei, Xiaokai and Huang, Wei-Chieh and Hui, Zheng and Wang, Chen and Gong, Michelle and Yu, Philip S.",
|
| 222 |
+
booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2026",
|
| 223 |
+
year = "2026",
|
| 224 |
+
pages = "38554--38575",
|
| 225 |
+
url = "https://aclanthology.org/2026.findings-emnlp.1804/",
|
| 226 |
+
doi = "10.18653/2026.findings-emnlp.1804"
|
| 227 |
+
}
|
| 228 |
+
```
|
| 229 |
|
| 230 |
+
Derived from [Amazon Reviews 2023](https://huggingface.co/datasets/McAuley-Lab/Amazon-Reviews-2023). Original reviews and images retain applicable source terms and third-party rights. This packaging does not grant new rights to those materials.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 231 |
|
| 232 |
+
Data release: **2026-11-12** · Negative-item view and documentation: **2026-11-18**. Shared tables are identical across URL and Full editions; Full additionally includes cached image bytes.
|
manifest.json
CHANGED
|
@@ -4,8 +4,8 @@
|
|
| 4 |
"files": [
|
| 5 |
{
|
| 6 |
"path": "README.md",
|
| 7 |
-
"bytes":
|
| 8 |
-
"sha256": "
|
| 9 |
},
|
| 10 |
{
|
| 11 |
"path": "evaluation/source_candidates/hard_negatives_Beauty_and_Personal_Care_v3.jsonl",
|
|
@@ -207,6 +207,11 @@
|
|
| 207 |
"bytes": 1743,
|
| 208 |
"sha256": "b33d349b41046d6e9e508733b3614b3ceb681d7693bdc07ffb53963cfee5b02d"
|
| 209 |
},
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 210 |
{
|
| 211 |
"path": "reports/release_validation.json",
|
| 212 |
"bytes": 3223,
|
|
@@ -332,6 +337,11 @@
|
|
| 332 |
"bytes": 5014826,
|
| 333 |
"sha256": "623d74e2f3259a822f2acd381fcd7d005baafb47ab57325baecee7e2acc6567b"
|
| 334 |
},
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 335 |
{
|
| 336 |
"path": "tables/products/All_Beauty.parquet",
|
| 337 |
"bytes": 180088,
|
|
|
|
| 4 |
"files": [
|
| 5 |
{
|
| 6 |
"path": "README.md",
|
| 7 |
+
"bytes": 12408,
|
| 8 |
+
"sha256": "608269d861a81b7782366a1c047ecb3c74d03e4f0f36aabb7af9d843ad7539ee"
|
| 9 |
},
|
| 10 |
{
|
| 11 |
"path": "evaluation/source_candidates/hard_negatives_Beauty_and_Personal_Care_v3.jsonl",
|
|
|
|
| 207 |
"bytes": 1743,
|
| 208 |
"sha256": "b33d349b41046d6e9e508733b3614b3ceb681d7693bdc07ffb53963cfee5b02d"
|
| 209 |
},
|
| 210 |
+
{
|
| 211 |
+
"path": "reports/negative_metadata_coverage.json",
|
| 212 |
+
"bytes": 763,
|
| 213 |
+
"sha256": "27a35444a286714b2cb49ba69bf05f5bad1424b7e0fa7ba0512e22d337ea2e8a"
|
| 214 |
+
},
|
| 215 |
{
|
| 216 |
"path": "reports/release_validation.json",
|
| 217 |
"bytes": 3223,
|
|
|
|
| 337 |
"bytes": 5014826,
|
| 338 |
"sha256": "623d74e2f3259a822f2acd381fcd7d005baafb47ab57325baecee7e2acc6567b"
|
| 339 |
},
|
| 340 |
+
{
|
| 341 |
+
"path": "tables/negative_items/test.parquet",
|
| 342 |
+
"bytes": 36411001,
|
| 343 |
+
"sha256": "d706bcf9c2e1e55b963f8fb11600f21d3b3fae7d4e0713d45fc7c77ca9a14fea"
|
| 344 |
+
},
|
| 345 |
{
|
| 346 |
"path": "tables/products/All_Beauty.parquet",
|
| 347 |
"bytes": 180088,
|
reports/negative_metadata_coverage.json
ADDED
|
@@ -0,0 +1,34 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"status": "passed",
|
| 3 |
+
"sampling": "Existing saved candidates; no resampling",
|
| 4 |
+
"candidate_rows": 19751,
|
| 5 |
+
"unique_negative_product_domain_pairs": 211364,
|
| 6 |
+
"negative_items_outside_history": 204915,
|
| 7 |
+
"missing_metadata": [],
|
| 8 |
+
"domains": [
|
| 9 |
+
{
|
| 10 |
+
"domain": "Beauty_and_Personal_Care",
|
| 11 |
+
"negative_items": 51966,
|
| 12 |
+
"outside_history": 49416
|
| 13 |
+
},
|
| 14 |
+
{
|
| 15 |
+
"domain": "Books",
|
| 16 |
+
"negative_items": 53433,
|
| 17 |
+
"outside_history": 52571
|
| 18 |
+
},
|
| 19 |
+
{
|
| 20 |
+
"domain": "Electronics",
|
| 21 |
+
"negative_items": 52641,
|
| 22 |
+
"outside_history": 51075
|
| 23 |
+
},
|
| 24 |
+
{
|
| 25 |
+
"domain": "Home_and_Kitchen",
|
| 26 |
+
"negative_items": 53324,
|
| 27 |
+
"outside_history": 51853
|
| 28 |
+
}
|
| 29 |
+
],
|
| 30 |
+
"tasks": {
|
| 31 |
+
"item_ranking": 9876,
|
| 32 |
+
"item_preference_choice": 9875
|
| 33 |
+
}
|
| 34 |
+
}
|
tables/negative_items/test.parquet
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:d706bcf9c2e1e55b963f8fb11600f21d3b3fae7d4e0713d45fc7c77ca9a14fea
|
| 3 |
+
size 36411001
|