alete-ai/web-cache-similarity-benchmark

Web Cache Semantic Similarity Benchmark

This dataset was created by Alete to benchmark the effectiveness of modern semantic vector caching against traditional cryptographic string hashing (SHA-256) directly on the live web, evaluating temporal cache decay across multiple time frames in a uniform environment.

Naïve string hashing fails on the modern web because minor changes in page layouts—such as dynamic trending articles list, dynamic comments count, ad units, and modified timestamps—alter the byte signature completely. This causes false negatives (cache misses), forcing backend LLM extraction pipelines to consume extra tokens and reprocess pages that have not fundamentally changed.

This benchmark evaluates 50 live articles across three snapshots ($T_1$, $T_2$, and $T_3$) fetched using the identical direct raw HTTP scrape configuration (requests.get with user-agent mimicking and BeautifulSoup parsing), compiled via MDream minimal preset:

• $T_1$ (Baseline): Initial direct HTTP fetch at 11:44:08-04:00.
• $T_2$ (3.5-Hour Snapshot): Direct HTTP fetch at 15:17:12-04:00.
• $T_3$ (5.5-Hour Snapshot): Direct HTTP fetch at 17:25:02-04:00.

---

Dataset Schema

Each row contains the following metrics and content snapshots:

Field Name	Type	Description
url	string	The live publication URL of the news article.
source	string	The news publisher source (fox, bbc, guardian).
timestamp_t1	string	Timestamp for the baseline direct HTTP crawl ($T_1$).
timestamp_t2	string	Timestamp for the direct HTTP crawl ($T_2$).
timestamp_t3	string	Timestamp for the direct HTTP crawl ($T_3$).
markdown_t1	string	Markdown compiled from raw HTML at $T_1$ using MDream minimal preset.
markdown_t2	string	Markdown compiled from raw HTML at $T_2$ using MDream minimal preset.
markdown_t3	string	Markdown compiled from raw HTML at $T_3$ using MDream minimal preset.
digest_t1	string	M-Dream AST digest (Title + Headings + Lead 800 chars) for $T_1$.
digest_t2	string	M-Dream AST digest for $T_2$.
digest_t3	string	M-Dream AST digest for $T_3$.
tokens_t3_digest	integer	Token footprint of the digest at $T_3$.
tokens_t3_full	integer	Token footprint of the full markdown body at $T_3$.
sha_full_match	boolean	True if SHA256(markdown_t1) == SHA256(markdown_t2) (3.5-hour HTTP match).
sha_digest_match	boolean	True if SHA256(digest_t1) == SHA256(digest_t2) (3.5-hour digest match).
digest_similarity	float	Cosine similarity score of $T_1$ vs $T_2$ digests using Vertex AI text-embedding-005.
full_similarity	float	Cosine similarity score of $T_1$ vs $T_2$ full markdown using text-embedding-005.
sha_t1_t3_match	boolean	True if SHA256(markdown_t1) == SHA256(markdown_t3) (5.5-hour HTTP match).
sha_t1_t3_digest_match	boolean	True if SHA256(digest_t1) == SHA256(digest_t3) (5.5-hour digest match).
t1_t3_digest_similarity	float	Cosine similarity score of $T_1$ vs $T_3$ digests.
t1_t3_full_similarity	float	Cosine similarity score of $T_1$ vs $T_3$ full markdown.
sha_t2_t3_match	boolean	True if SHA256(markdown_t2) == SHA256(markdown_t3) (2-hour HTTP match).
sha_t2_t3_digest_match	boolean	True if SHA256(digest_t2) == SHA256(digest_t3) (2-hour digest match).
t2_t3_digest_similarity	float	Cosine similarity score of $T_2$ vs $T_3$ digests.
t2_t3_full_similarity	float	Cosine similarity score of $T_2$ vs $T_3$ full markdown.

---

3-Snapshot Cache Benchmark Results

The empirical results demonstrate the performance of caching under same-engine direct temporal decay:

1. 5.5-Hour Temporal Decay ($T_1$ vs $T_3$: HTTP-to-HTTP)

• SHA-256 Full Markdown Hits: 20.0% (80.0% cache misses due to temporal dynamic sidebar adjustments).
• SHA-256 M-Dream Digest Hits: 92.0% (AST container isolation blocks most page noise).
• Digest Embedding Hits (threshold 0.92): 98.0% (Consistent semantic cache hit rate).
• Full-Text Embedding Hits (threshold 0.92): 98.0%.

2. 2.0-Hour Temporal Decay ($T_2$ vs $T_3$: HTTP-to-HTTP)

• SHA-256 Full Markdown Hits: 28.0% (72.0% cache misses even after just 2 hours!).
• SHA-256 M-Dream Digest Hits: 100.0% (M-Dream digest achieves a perfect 100.0% cache hit rate over 2 hours).
• Digest Embedding Hits (threshold 0.92): 100.0%.
• Full-Text Embedding Hits (threshold 0.92): 100.0%.

---

How to Use the Dataset

You can load this dataset programmatically using the Hugging Face datasets library:

from datasets import load_dataset

dataset = load_dataset("alete-ai/web-cache-similarity-benchmark")
print(dataset["train"][0])