Datasets:

kylebrodeur
/

embedding-eval-results

name stringclasses 7 values	provider stringclasses 2 values	model stringclasses 7 values	dim int64 384 1.02k	recall@1 float64 0.48 0.83	recall@5 float64 0.7 1	recall@10 float64 0.76 1	mrr@10 float64 0.58 0.9	q_latency_ms float64 6.2 34.6
embeddinggemma	ollama	embeddinggemma	768	0.6552	0.9655	1	0.8027	18.2
bge-m3	ollama	bge-m3	1,024	0.6897	1	1	0.8	31.5
bge-large	ollama	bge-large	1,024	0.8276	0.931	0.9655	0.8693	29.4
nomic-embed-text	ollama	nomic-embed-text	768	0.8276	0.9655	1	0.9004	16.1
mxbai-embed-large	ollama	mxbai-embed-large	1,024	0.7931	0.931	0.9655	0.8487	27
paraphrase-multilingual	ollama	paraphrase-multilingual	768	0.6552	0.931	1	0.7869	16.4
all-minilm (current local fallback)	ollama	all-minilm	384	0.6552	0.9655	1	0.7787	8.3
embeddinggemma	ollama	embeddinggemma	768	0.8	0.92	0.9267	0.8524	18.7
bge-m3	ollama	bge-m3	1,024	0.7533	0.9067	0.9333	0.8172	34.6
bge-large	ollama	bge-large	1,024	0.7333	0.88	0.9133	0.7984	28.3
nomic-embed-text	ollama	nomic-embed-text	768	0.7267	0.86	0.9	0.7912	14.8
mxbai-embed-large	ollama	mxbai-embed-large	1,024	0.7067	0.8933	0.9067	0.7865	29.5
paraphrase-multilingual	ollama	paraphrase-multilingual	768	0.62	0.78	0.8333	0.6903	16.3
all-minilm (current local fallback)	ollama	all-minilm	384	0.64	0.7933	0.8333	0.6998	6.2
embeddinggemma	ollama	embeddinggemma	768	0.7089	0.9022	0.94	0.7918	20.2
bge-m3	ollama	bge-m3	1,024	0.68	0.88	0.92	0.7669	32
bge-large	ollama	bge-large	1,024	0.6111	0.8267	0.8756	0.7056	27.7
nomic-embed-text	ollama	nomic-embed-text	768	0.6156	0.8511	0.9	0.716	13.7
mxbai-embed-large	ollama	mxbai-embed-large	1,024	0.6356	0.8444	0.8956	0.7303	28
paraphrase-multilingual	ollama	paraphrase-multilingual	768	0.4844	0.7	0.76	0.5773	16.4
all-minilm (current local fallback)	ollama	all-minilm	384	0.5911	0.7978	0.8489	0.6841	6.5
embeddinggemma	modal	embeddinggemma	768	0.6533	0.8889	0.9356	0.7524	6.7

Embedding Eval Results

The committed results from a real embedding-model benchmark that embarrassed a leaderboard's recommendation.

What's in here

Four CSV files representing four evaluation runs on a personal Obsidian vault:

File	Notes	Queries	Models	Purpose
`20260618-233912.csv`	54	29	7	Round 1 — toy slice. Saturated benchmark.
`20260619-025330.csv`	995	150	7	Stage 1 — real pile. The ranking inverted.
`20260620-142445-wholenote.csv`	996	450	7	Stage 1 enlarged — held the new order.
`20260620-172625-chunked.csv`	18,279 chunks	450	1	Chunked A/B — chunking hurt every metric.

What the data shows

Round 1 on 54 notes: all 7 models looked great. recall@5 above 90% for every candidate. The "winner" by leaderboard-style ranking was nomic-embed-text.

Stage 1 on 995 notes: the ranking inverted. The mid-pack model embeddinggemma took first on every metric. The round-one "winner" sank to mid-pack. The default the system had been quietly using, all-minilm, dropped to last.

Same code. Same models. Same scoring. The only thing that changed was the size of the pile.

Columns

Each CSV has the same columns:

name,provider,model,dim,recall@1,recall@5,recall@10,mrr@10,q_latency_ms,approx_cost_usd

name: human-readable label (matches the model except for the legacy all-minilm row, which is annotated)
provider: ollama (local) or modal (the project's own GPU service) — see sources below
model: the actual model name passed to the embedder
dim: embedding dimension
recall@1, recall@5, recall@10: fraction of queries where the source note landed in the top-k
mrr@10: mean reciprocal rank
q_latency_ms: query-side embedding latency (note embedding happens separately)
approx_cost_usd: zero for local models; the cheapest paid model cost roughly a few cents per run

Source

The full eval harness is in the pi-vault-mind repo at eval/run_eval.py. The companion blog post is "Pick the Model From Your Own Data" (KE-1).

The simplified standalone version of the harness, written for the blog post, is at kylebrodeur/embed-eval-on-your-vault. It is a single-file Python script with no dependencies beyond python3 that runs against any vault of .md or .txt notes.

Reproducibility

Each CSV is timestamped (UTC). The harness commits the result on every run. Re-running with the same model/provider/dataset should produce the same numbers within rounding. Embedding models do drift across versions, so if you re-run after a model bump, expect ±1-2% on the metrics.

Citation

If you use these numbers in another piece, cite the model row and the run timestamp. The full provenance is in the pi-vault-mind repo commit log (the 2026-06-18 to 2026-06-20 range).

License

MIT. Numbers are facts; the framing is the author's.

Downloads last month: 17

Number of rows:

Total file size:

7.95 kB