Datasets:

exa-ai
/

ATLAS

---
license: mit
language:
- en
pretty_name: ATLAS
size_categories:
- n<1K
task_categories:
- question-answering
- table-question-answering
tags:
- benchmark
- agent
- agentic-search
- web-search
- deep-research
- information-retrieval
- factuality
- llm-evaluation
configs:
- config_name: 2026-10
  default: true
  data_files:
  - split: test
    path: atlas-2026-10.jsonl
extra_gated_heading: Request access to ATLAS
extra_gated_prompt: >-
  ATLAS is an eval set. Please don't post its questions or answers online in plain text.
extra_gated_fields:
  Affiliation: text
extra_gated_button_content: Request access
---
# ATLAS
#### 547 deep and wide research tasks that grade the accuracy and completeness of search agents.

**Access is limited for now.** We're sharing ATLAS with a small group first, so we're approving only a small number of requests
and may not be able to reply to each one. We plan to open access more widely over time, and the
[grader code](https://github.com/exa-labs/benchmarks/tree/master/benchmarks/graders/atlas) is public in the meantime. Thank you for your
interest and your patience.

📝 [Blog post](https://exa.ai/blog/atlas-benchmark)\
💻 [Evaluation code](https://github.com/exa-labs/benchmarks/tree/master/benchmarks/graders/atlas)\
🐛 [Report an issue](https://github.com/exa-labs/benchmarks/issues)

## Benchmark

ATLAS (Agentic Tasks for Large Aggregation + Search) is a benchmark designed to grade the
accuracy and completeness of agent responses to challenging real-world workflows that
heavily utilize web search. It pairs queries grounded
in real search demand with verified golden answers, using an automated pipeline that lets us
refresh both as the web and models evolve.

Each task asks a system to discover every entity (e.g. a person, place or company) that
meets precise conditions, and then to enrich each entity with 2–10 multi-hop attributes.
Tasks are generated from seed topics drawn from clustering of anonymized search demand.

A committee of frontier models and search providers spends a median of 8 agent-hours and
1,200 searches per task to find answers, independently verify them against sources, and
resolve disagreements or ambiguities. Blind judges then research every disagreement between
the golden answers and twelve search systems, and Exa engineers manually adjudicate anything
still in dispute. We estimate that under 1% of gold values are incorrect (0.9%, 95% CI
0.6–1.4%).

The challenge ATLAS poses is efficient search: solve each task in under a minute and for
under $1, a budget at which no current system reaches even 0.5 row F1.

## Example

Here is what an ATLAS task looks like. So that no real task appears online in plain text,
this example comes from a separate test run of our pipeline and is not one of the 547 tasks.

**Question** (abridged; the full task also defines every column and the edge cases):

> Which Legend of Zelda video games have been adapted into a reworked version (a product
> marketed or built as a "remake," "remaster," "HD," "3D," or "definitive/anniversary
> edition") that actually went on sale on or before 11 October 2026, published and sold by
> Nintendo for the Nintendo 3DS, Wii U, Nintendo Switch or Nintendo Switch 2? …

**Golden table** (4 of its 14 rows):

| Zelda remake/remaster product title | Critic Metascore | Lead developer founding year | Lead developer headquarters city | Original game's U.S. age rating | Console Japan launch MSRP (JPY) | Hero voice actor birth year | Hero voice actor home prefecture |
|---|---|---|---|---|---|---|---|
| The Legend of Zelda: Ocarina of Time 3D | 94 | 2006 | Tokyo | E | 25000 | 1967 | Hiroshima |
| The Legend of Zelda: The Wind Waker HD | 90 | 1889 | Kyoto | E | 26250 | 1973 | Kyoto |
| The Legend of Zelda: Twilight Princess HD | 86 | 1994 | Melbourne | T | 26250 | 1973 | Tokyo |
| The Legend of Zelda: Skyward Sword HD | 81 | 1994 | Melbourne | E10+ | 32378 | 1978 | Kanagawa |

## Dataset Description

547 examples, each with:

* `id`: the task id.
* `question`: the task, including the columns to return and the answer format (encrypted).
* `gold`: the golden table, plus the grading rule for each column (encrypted).
* `canary`: the canary string, which is also the decryption key.

As with BrowseComp, the questions and golden tables are encrypted so they never appear
online as plain text. To read them:

```python
import base64, hashlib, json
from datasets import load_dataset
def decrypt(text, canary):
    data = base64.b64decode(text)
    digest = hashlib.sha256(canary.encode()).digest()
    key = (digest * (len(data) // len(digest) + 1))[: len(data)]
    return bytes(a ^ b for a, b in zip(data, key)).decode()
atlas = load_dataset("exa-ai/ATLAS", split="test")
task = atlas[0]
question = decrypt(task["question"], task["canary"])
gold = json.loads(decrypt(task["gold"], task["canary"]))
```

## Evaluation

The grader is in
[exa-labs/benchmarks](https://github.com/exa-labs/benchmarks/tree/master/benchmarks/graders/atlas).
To run it:

```bash
uv run bench run --suite atlas --system scout-exa-auto-highlights
```

We grade returned tables against golden answers using three F1 metrics, computed per task
and averaged over tasks:

* **Discovery F1** counts just row discovery: how many returned rows name a gold entity, and
  how many gold entities are returned.
* **Item F1** counts cells, including both discovery and enrichment.
* **Row F1** counts rows, where a row is correct only when every cell in the row is. This is
  our headline metric.

The grader aligns returned rows to the golden table by entity name and aliases, falling back
to an LLM aligner (GPT-6 Luna) on misses; different versions of a model, product or software
count as different entities. It then scores each cell deterministically where
possible, with an LLM judge (GPT-6 Luna) adjudicating the remaining 4.5% of cells. Blank
cells, where our agents could show that no data exists, earn credit only when left empty,
while ungraded cells, which could not be settled, are excluded. Grading one system on all
547 tasks costs ~$0.24.

## Limitations

ATLAS is synthetically generated: frontier-model agents built and verified every task and
golden table, and we estimate that under 1% of gold values are incorrect. The golden tables
reflect the web as of late September 2026. Facts change, so some cells will go out of date.

If you find a clear error, please [open an issue](https://github.com/exa-labs/benchmarks/issues)
with the task id, the row and column, and a link to a source that shows the correct value.
Please don't paste the question or the golden table into the issue. We fix confirmed errors
in tagged revisions and plan to regenerate ATLAS from fresh demand every 3–6 months.

## Citation

```bibtex
@online{exa2026atlas,
  author = {Alexander Goldberg and Joshua Ahn and Scott Langille},
  title  = {{ATLAS: Evaluating Agents on Search-Intensive Tasks}},
  date   = {2026-10-08},
  url    = {https://exa.ai/blog/atlas-benchmark}
}
```

	---
	license: mit
	language:
	- en
	pretty_name: ATLAS
	size_categories:
	- n<1K
	task_categories:
	- question-answering
	- table-question-answering
	tags:
	- benchmark
	- agent
	- agentic-search
	- web-search
	- deep-research
	- information-retrieval
	- factuality
	- llm-evaluation
	configs:
	- config_name: 2026-10
	default: true
	data_files:
	- split: test
	path: atlas-2026-10.jsonl
	extra_gated_heading: Request access to ATLAS
	extra_gated_prompt: >-
	ATLAS is an eval set. Please don't post its questions or answers online in plain text.
	extra_gated_fields:
	Affiliation: text
	extra_gated_button_content: Request access
	---

	# ATLAS
	#### 547 deep and wide research tasks that grade the accuracy and completeness of search agents.

	Access is limited for now. We're sharing ATLAS with a small group first, so we're approving only a small number of requests
	and may not be able to reply to each one. We plan to open access more widely over time, and the
	[grader code](https://github.com/exa-labs/benchmarks/tree/master/benchmarks/graders/atlas) is public in the meantime. Thank you for your
	interest and your patience.

	📝 [Blog post](https://exa.ai/blog/atlas-benchmark)\
	💻 [Evaluation code](https://github.com/exa-labs/benchmarks/tree/master/benchmarks/graders/atlas)\
	🐛 [Report an issue](https://github.com/exa-labs/benchmarks/issues)

	## Benchmark

	ATLAS (Agentic Tasks for Large Aggregation + Search) is a benchmark designed to grade the
	accuracy and completeness of agent responses to challenging real-world workflows that
	heavily utilize web search. It pairs queries grounded
	in real search demand with verified golden answers, using an automated pipeline that lets us
	refresh both as the web and models evolve.

	Each task asks a system to discover every entity (e.g. a person, place or company) that
	meets precise conditions, and then to enrich each entity with 2–10 multi-hop attributes.
	Tasks are generated from seed topics drawn from clustering of anonymized search demand.

	A committee of frontier models and search providers spends a median of 8 agent-hours and
	1,200 searches per task to find answers, independently verify them against sources, and
	resolve disagreements or ambiguities. Blind judges then research every disagreement between
	the golden answers and twelve search systems, and Exa engineers manually adjudicate anything
	still in dispute. We estimate that under 1% of gold values are incorrect (0.9%, 95% CI
	0.6–1.4%).

	The challenge ATLAS poses is efficient search: solve each task in under a minute and for
	under $1, a budget at which no current system reaches even 0.5 row F1.

	## Example

	Here is what an ATLAS task looks like. So that no real task appears online in plain text,
	this example comes from a separate test run of our pipeline and is not one of the 547 tasks.

	Question (abridged; the full task also defines every column and the edge cases):

	> Which Legend of Zelda video games have been adapted into a reworked version (a product
	> marketed or built as a "remake," "remaster," "HD," "3D," or "definitive/anniversary
	> edition") that actually went on sale on or before 11 October 2026, published and sold by
	> Nintendo for the Nintendo 3DS, Wii U, Nintendo Switch or Nintendo Switch 2? …

	Golden table (4 of its 14 rows):

	\| Zelda remake/remaster product title \| Critic Metascore \| Lead developer founding year \| Lead developer headquarters city \| Original game's U.S. age rating \| Console Japan launch MSRP (JPY) \| Hero voice actor birth year \| Hero voice actor home prefecture \|
	\|---\|---\|---\|---\|---\|---\|---\|---\|
	\| The Legend of Zelda: Ocarina of Time 3D \| 94 \| 2006 \| Tokyo \| E \| 25000 \| 1967 \| Hiroshima \|
	\| The Legend of Zelda: The Wind Waker HD \| 90 \| 1889 \| Kyoto \| E \| 26250 \| 1973 \| Kyoto \|
	\| The Legend of Zelda: Twilight Princess HD \| 86 \| 1994 \| Melbourne \| T \| 26250 \| 1973 \| Tokyo \|
	\| The Legend of Zelda: Skyward Sword HD \| 81 \| 1994 \| Melbourne \| E10+ \| 32378 \| 1978 \| Kanagawa \|

	## Dataset Description

	547 examples, each with:

	* `id`: the task id.
	* `question`: the task, including the columns to return and the answer format (encrypted).
	* `gold`: the golden table, plus the grading rule for each column (encrypted).
	* `canary`: the canary string, which is also the decryption key.

	As with BrowseComp, the questions and golden tables are encrypted so they never appear
	online as plain text. To read them:

	```python
	import base64, hashlib, json
	from datasets import load_dataset

	def decrypt(text, canary):
	data = base64.b64decode(text)
	digest = hashlib.sha256(canary.encode()).digest()
	key = (digest * (len(data) // len(digest) + 1))[: len(data)]
	return bytes(a ^ b for a, b in zip(data, key)).decode()

	atlas = load_dataset("exa-ai/ATLAS", split="test")
	task = atlas[0]
	question = decrypt(task["question"], task["canary"])
	gold = json.loads(decrypt(task["gold"], task["canary"]))
	```

	## Evaluation

	The grader is in
	[exa-labs/benchmarks](https://github.com/exa-labs/benchmarks/tree/master/benchmarks/graders/atlas).
	To run it:

	```bash
	uv run bench run --suite atlas --system scout-exa-auto-highlights
	```

	We grade returned tables against golden answers using three F1 metrics, computed per task
	and averaged over tasks:

	* Discovery F1 counts just row discovery: how many returned rows name a gold entity, and
	how many gold entities are returned.
	* Item F1 counts cells, including both discovery and enrichment.
	* Row F1 counts rows, where a row is correct only when every cell in the row is. This is
	our headline metric.

	The grader aligns returned rows to the golden table by entity name and aliases, falling back
	to an LLM aligner (GPT-6 Luna) on misses; different versions of a model, product or software
	count as different entities. It then scores each cell deterministically where
	possible, with an LLM judge (GPT-6 Luna) adjudicating the remaining 4.5% of cells. Blank
	cells, where our agents could show that no data exists, earn credit only when left empty,
	while ungraded cells, which could not be settled, are excluded. Grading one system on all
	547 tasks costs ~$0.24.

	## Limitations

	ATLAS is synthetically generated: frontier-model agents built and verified every task and
	golden table, and we estimate that under 1% of gold values are incorrect. The golden tables
	reflect the web as of late September 2026. Facts change, so some cells will go out of date.

	If you find a clear error, please [open an issue](https://github.com/exa-labs/benchmarks/issues)
	with the task id, the row and column, and a link to a source that shows the correct value.
	Please don't paste the question or the golden table into the issue. We fix confirmed errors
	in tagged revisions and plan to regenerate ATLAS from fresh demand every 3–6 months.

	## Citation

	```bibtex
	@online{exa2026atlas,
	author = {Alexander Goldberg and Joshua Ahn and Scott Langille},
	title = {{ATLAS: Evaluating Agents on Search-Intensive Tasks}},
	date = {2026-10-08},
	url = {https://exa.ai/blog/atlas-benchmark}
	}
	```