Datasets:

zc277584121
/

modern-embedding-bench

text_a stringlengths 16 215	text_b stringlengths 13 199	score float64 0 5
A girl is styling her hair.	A girl is brushing her hair.	2.5
A group of men play soccer on the beach.	A group of boys are playing soccer on the beach.	3.6
One woman is measuring another woman's ankle.	A woman measures another woman's ankle.	5
A man is cutting up a cucumber.	A man is slicing a cucumber.	4.2
A man is playing a harp.	A man is playing a keyboard.	1.5
A woman is cutting onions.	A woman is cutting tofu.	1.8
A man is riding an electric bicycle.	A man is riding a bicycle.	3.5
A man is playing the drums.	A man is playing the guitar.	2.2
A man is playing guitar.	A lady is playing the guitar.	2.2
A man is playing a guitar.	A man is playing a trumpet.	1.71
A man is playing a guitar.	A man is playing a trumpet.	1.71
A man is cutting an onion.	A man cuts an onion.	5
A man is cycling.	A man is talking.	0.6
A man is slicing open a fish.	A man is cutting up a fish.	4.4
A man is slicing a tomato.	A man is slicing a bun.	2
A man is playing a guitar.	A man is playing a keyboard.	1.8
A baby panda goes down a slide.	A panda slides down a slide.	4.4
A man is singing and playing a guitar.	A man is playing a guitar.	3.6
A man attacks a woman.	A man slaps a woman.	3.6
A man is driving a car.	A man is riding a horse.	1.2
A woman is cutting tofu.	A woman is cutting an onion.	2.4
The woman is styling her hair.	The woman is slicing herbs.	0.2
Two zebras play in an open field.	Two zebras are playing in a field.	4.2
A man is cutting a potato.	A man is slicing some potato.	4.4
A man is slicing an onion.	A woman is slicing a pumpkin.	2.25
A man is dancing.	A man and woman is dancing.	2
A man is riding a motorcycle.	A woman is riding a horse.	0.75
A woman is slicing garlics.	A woman is slicing an onion.	2.2
A man is speaking.	A man is cooking.	0.8
A little boy is singing and playing a guitar.	A man is singing and playing the guitar.	2.2
A turtle is swimming in water.	A turtle is walking underwater.	3.2
A young woman is putting stickers all over her face.	A woman is applying stickers to her face.	4.8
A woman is wrapping tofu.	A woman is balling dough.	1.4
A cat is eating some corn.	A cat is eating corn on the cob.	4.25
A man is eating a food.	A man is eating a piece of bread.	3.4
A man is playing a guitar.	A man is eating pasta.	0.53
A man is kicking pots of water.	A man is picking flowers.	0.4
A man is cutting a pipe with scissors.	A man is cutting carpet with a knife.	1.2
A woman is dancing in the rain.	A woman dances in the rain out side.	5
A woman is taking a bath.	A woman is riding a horse.	0.54
A man mixes vegetables in a pot.	A person is stirring vegetables in a pot.	3.75
A woman is talking on a cell phone.	A man and woman are talking on the phone.	3
A man is playing a guitar.	A man is singing while playing the guitar.	3.6
A man is playing a guitar.	A man is driving a car.	0.5
A man is cutting apple by his hand.	A man is cutting carpet with a knife.	1.5
A man is opening a door.	A man is cutting an onion.	0.8
A man is slicing a tomato.	A man is riding a horse.	0.8
A man is cutting paper with a sword.	A woman is cutting a tomato.	0.6
A boy studies a calendar.	A boy is looking at a calendar.	4.4
The ballerina is dancing.	A man is dancing.	1.75
A woman is dancing.	A woman is playing violin.	0.4
A woman is slicing some tomatoes.	A woman is chopping a potato.	1.4
A woman is water skiing.	A woman is slicing fish.	0.4
A man is playing a flute.	A man is riding a scooter.	0.8
A man is playing the piano.	A man played the guitar.	2
A woman is picking a can.	A man is playing a guitar.	0.13
A man puts three pieces of meat into a pan.	A man is putting meat in a pan.	4
A woman is cutting an onion.	A woman is cleaning a garden.	0.27
Some men are sawing.	Men are sawing logs.	3.4
A car is driven down the road.	A girl is walking down a road.	1.2
The man is kissing and hugging the woman.	A man is hugging and kissing a woman.	5
A train is moving.	A man is doing yoga.	0
Someone is slicing an onion.	A woman is cutting onion.	3.8
A woman is playing in the ocean.	A woman is preparing shrimp to cook.	0.75
A person is playing an electronic keyboard.	A kid is playing keyboard.	3.4
A man is holding a leaf.	A monkey is fighting a man.	0
A woman is peeling shrimp.	A man is squeezing water.	0.2
A man is sitting and smoking.	A man is smoking a cigarette.	4
A man is playing a guitar.	A woman is riding a horse.	0.5
A man is standing in front of the window and looking outside.	A man is staring out the window.	3.8
A skunk is looking here and there.	A skunk looks at the camera.	2.4
A man is playing the guitar and singing.	A man sings with a guitar.	4.75
A woman opens a window.	A man is crawling.	0
People are dancing outside.	A group of people are dancing.	3.75
The man is using a camera to hammer a nail.	Someone is banging a camera lense against a nail.	2.6
A woman is filing her nails.	A man is peeling a carrot.	0
A boy is crawling into a dog house.	A boy is playing a wooden flute.	0.75
A woman is swimming underwater.	A man is slicing some carrots.	0
A machine is sharpening a pencil.	The machine shaved the end of the pencil.	3.8
A monkey is playing drums.	A gorilla plays the drums.	2.8
A man is opening a box and taking out paper.	A woman is peeling a potato.	0
A woman is dancing.	A woman plays the clarinet.	0.8
A person is drawing on a large touchscreen.	A man is drawing on a digital dry erase board.	3
The men played follow the leader on the grass.	The rhino grazed on the grass.	1
A woman is cracking eggs.	A man is talking to a woman.	0
A woman peels garlic with her hands.	The woman is slicing herbs.	1
The polar bears fought over the kill.	Polar bears are fighting each other.	3.4
A man is doing trick with play cards.	A man is performing a card trick.	5
The cat is licking a bottle.	A cat plays with a small bottle.	2.33
A person is slicing an onion.	A person cuts ginger.	1.4
A person is peeling a potato with a potato peeler.	A man is cutting tomatoes with a cleaver.	0.75
Two women are dancing and singing in front of a crowd.	The women are singing and dancing.	3.54
A man is seasoning some carrots.	A woman is slicing garlic.	0.8
Two men pushed carts through the woods.	Two men are pushing carts.	3.5
A man is playing a football.	A man is maneuvering a soccer ball with his feet.	2
The lady peeled the potatoe.	A woman is peeling a potato.	4.75
A woman is slicing some tofu.	A woman is cutting a block of tofu into small cubes.	4
Someone typed on a keyboard.	Someone is typing.	4.5
Three young men run, jump, and kick off of a Coke machine.	Three men are jumping off a wall.	1.5
A young Asian girl is applying eyeliner.	A girl is putting on eye makeup.	2.4

End of preview. Expand in Data Studio

Modern Embedding Bench

Modern Embedding Bench evaluates embedding models on practical retrieval tasks that show up in current AI systems but are often under-covered by broad leaderboards. The focus is on agent memory, tool and document retrieval, long-context RAG, cross-lingual technical retrieval, coding-oriented retrieval, and multimodal search rather than a single aggregate score.

The companion leaderboard Space is available at: https://huggingface.co/spaces/zc277584121/modern-embedding-bench-leaderboard

The source code is available at: https://github.com/zc277584121/modern-embedding-bench

models.jsonl: registered model specs
tasks.jsonl: registered task specs and primary metrics
runs/: run manifests used to produce results
results/latest.jsonl: canonical v2 result records
leaderboards/latest.csv: flat leaderboard table derived from result records
benchmark_data/: optional benchmark input data exported from the local repo

Current Public Export

Registry model specs: 20
Public model specs exported: 18
Excluded private or preview model specs: 2
Task specs: 6
Result records: 292 records, 246 successful, 46 failed
Leaderboard rows: 239
Tasks with leaderboard rows: 4
Providers with leaderboard rows: 10
Unique task/model leaderboard pairs: 60
Duplicate task/model repeats kept for inspection: 179
Latest task/model marker rows: 60
Evidence tiers: legacy=239
Data: Bundled JSONL benchmark inputs are included.

Tasks

ID	Name	Primary metric	What it probes
autonomous_driving	Autonomous driving retrieval	avg_recall@1	Scenario retrieval for autonomous-driving style multimodal/text fallback cases.
chinese_multimodal	Chinese multimodal retrieval	avg_recall@1	Chinese multimodal/text fallback retrieval scenarios.
cross_modal_retrieval	Text-image retrieval	hard_avg_recall@1	COCO-style text-image matching with hard negative captions.
crosslingual_retrieval	Chinese-English retrieval	hard_avg_recall@1	Bidirectional technical retrieval with hard negatives across Chinese and English.
mrl_stress	MRL compression stress	spearman_dim_128	Semantic stability when embeddings are truncated to smaller dimensions.
needle_in_haystack	Long-document needle retrieval	overall_accuracy	Retrieving facts inserted at different positions in long documents.

Result Format

Each line in results/latest.jsonl is one model-task run. Important fields:

run: run id, description, metadata, and git sha when available
model: model id, display name, provider, modalities, dimensions, and tags
task: task id, dataset version, primary metric, and task kwargs
metrics: task-specific metric dictionary
details: diagnostic details for deeper analysis
error: error text for failed runs, otherwise null

Leaderboard Provenance

leaderboards/latest.csv keeps every public row, including historical duplicate runs for the same task_id and model_id. The first columns remain compatible with older CSV readers, and provenance columns are appended:

evidence_tier: legacy, smoke, benchmark, or unknown
evidence_source: legacy source file, git sha, or run id when available
task_model_duplicate_count: rows kept for the same task/model pair
task_model_run_rank: 1-based order for that task/model pair
is_latest_for_task_model: true for the latest exported row in that pair

Latest markers are computed from the order of results/latest.jsonl when result records are available, otherwise from CSV row order. Use is_latest_for_task_model=true to inspect one current row per task/model pair without losing the full historical trail.

Usage

Install and inspect the registry:

uv sync
uv run modern-embed-bench benchmark models
uv run modern-embed-bench benchmark tasks

Run a small OpenAI smoke benchmark:

uv run modern-embed-bench benchmark run \
  --manifest benchmark/runs/openai-smoke.yaml \
  --output results/openai-smoke.jsonl \
  --overwrite

uv run modern-embed-bench benchmark leaderboard \
  --results results/openai-smoke.jsonl \
  --output results/openai-smoke-leaderboard.csv

Notes and Limitations

Rows imported from legacy runs are published for continuity and should be read as historical baseline evidence, not as a fully normalized one-shot run.
Scores are task-specific. Avoid comparing scores across tasks as if they were one global ranking.
Some preview or private-in-progress model results are intentionally excluded from the public export until they are ready for publication.
Image binaries are not bundled by default; cross_modal metadata is included separately from the source image files.

Downloads last month: 60