strata-bench 1.1.0__tar.gz

strata_bench-1.1.0/.gitignore

@@ -0,0 +1,19 @@
+__pycache__/
+*.py[cod]
+*.egg-info/
+.eggs/
+dist/
+build/
+.pytest_cache/
+.ruff_cache/
+.venv/
+venv/
+.env
+.DS_Store
+.idea/
+.vscode/
+*.orig
+.coverage
+htmlcov/
+submissions/private/
+data/hidden/gold.json.bak

strata_bench-1.1.0/CITATION.cff

@@ -0,0 +1,30 @@
+cff-version: 1.2.0
+title: "STRATA-Bench: Evaluating AI Agents on Fragmented Spatial-Temporal Market Intelligence"
+message: If you use STRATA-Bench, please cite it as below.
+type: software
+authors:
+  - family-names: Movahedi
+    given-names: Mohammad
+    email: movahedi.m@northeastern.edu
+    affiliation: "The Globe and Mail / Northeastern University"
+    website: https://movahedi.ca
+repository-code: https://github.com/movahedi-ca/strata-bench
+url: https://github.com/movahedi-ca/strata-bench
+license: Apache-2.0
+version: 1.1.0
+date-released: "2026-09-19"
+keywords:
+  - AI agents
+  - benchmark
+  - retrieval
+  - spatial data
+  - hallucination
+  - real estate
+  - evaluation
+abstract: >-
+  STRATA-Bench (Spatial-Temporal Retrieval, Alignment, Transparency & Audit)
+  is a 39-task benchmark that measures whether AI agents can compile
+  longitudinal market intelligence when public data is geographically nested,
+  temporally gappy, and unit-inconsistent — without relabeling metro figures
+  as downtown, silently interpolating, or plotting missing quarters on a
+  categorical axis.

strata_bench-1.1.0/LICENSE

@@ -0,0 +1,190 @@
+                                 Apache License
+                           Version 2.0, January 2004
+                        http://www.apache.org/licenses/
+
+TERMS AND CONDITIONS FOR USE, REPRODUCTION, AND DISTRIBUTION
+
+1. Definitions.
+
+   "License" shall mean the terms and conditions for use, reproduction,
+   and distribution as defined by Sections 1 through 9 of this document.
+
+   "Licensor" shall mean the copyright owner or entity authorized by
+   the copyright owner that is granting the License.
+
+   "Legal Entity" shall mean the union of the acting entity and all
+   other entities that control, are controlled by, or are under common
+   control with that entity. For the purposes of this definition,
+   "control" means (i) the power, direct or indirect, to cause the
+   direction or management of such entity, whether by contract or
+   otherwise, or (ii) ownership of fifty percent (50%) or more of the
+   outstanding shares, or (iii) beneficial ownership of such entity.
+
+   "You" (or "Your") shall mean an individual or Legal Entity
+   exercising permissions granted by this License.
+
+   "Source" form shall mean the preferred form for making modifications,
+   including but not limited to software source code, documentation
+   source, and configuration files.
+
+   "Object" form shall mean any form resulting from mechanical
+   transformation or translation of a Source form, including but
+   not limited to compiled object code, generated documentation,
+   and conversions to other media types.
+
+   "Work" shall mean the work of authorship, whether in Source or
+   Object form, made available under the License, as indicated by a
+   copyright notice that is included in or attached to the work
+   (an example is provided in the Appendix below).
+
+   "Derivative Works" shall mean any work, whether in Source or Object
+   form, that is based on (or derived from) the Work and for which the
+   editorial revisions, annotations, elaborations, or other modifications
+   represent, as a whole, an original work of authorship. For the purposes
+   of this License, Derivative Works shall not include works that remain
+   separable from, or merely link (or bind by name) to the interfaces of,
+   the Work and Derivative Works thereof.
+
+   "Contribution" shall mean any work of authorship, including
+   the original version of the Work and any modifications or additions
+   to that Work or Derivative Works thereof, that is intentionally
+   submitted to Licensor for inclusion in the Work by the copyright owner
+   or by an individual or Legal Entity authorized to submit on behalf of
+   the copyright owner. For the purposes of this definition, "submitted"
+   means any form of electronic, verbal, or written communication sent
+   to the Licensor or its representatives, including but not limited to
+   communication on electronic mailing lists, source code control systems,
+   and issue tracking systems that are managed by, or on behalf of, the
+   Licensor for the purpose of discussing and improving the Work, but
+   excluding communication that is conspicuously marked or otherwise
+   designated in writing by the copyright owner as "Not a Contribution."
+
+   "Contributor" shall mean Licensor and any individual or Legal Entity
+   on behalf of whom a Contribution has been received by Licensor and
+   subsequently incorporated within the Work.
+
+2. Grant of Copyright License. Subject to the terms and conditions of
+   this License, each Contributor hereby grants to You a perpetual,
+   worldwide, non-exclusive, no-charge, royalty-free, irrevocable
+   copyright license to reproduce, prepare Derivative Works of,
+   publicly display, publicly perform, sublicense, and distribute the
+   Work and such Derivative Works in Source or Object form.
+
+3. Grant of Patent License. Subject to the terms and conditions of
+   this License, each Contributor hereby grants to You a perpetual,
+   worldwide, non-exclusive, no-charge, royalty-free, irrevocable
+   (except as stated in this section) patent license to make, have made,
+   use, offer to sell, sell, import, and otherwise transfer the Work,
+   where such license applies only to those patent claims licensable
+   by such Contributor that are necessarily infringed by their
+   Contribution(s) alone or by combination of their Contribution(s)
+   with the Work to which such Contribution(s) was submitted. If You
+   institute patent litigation against any entity (including a
+   cross-claim or counterclaim in a lawsuit) alleging that the Work
+   or a Contribution incorporated within the Work constitutes direct
+   or contributory patent infringement, then any patent licenses
+   granted to You under this License for that Work shall terminate
+   as of the date such litigation is filed.
+
+4. Redistribution. You may reproduce and distribute copies of the
+   Work or Derivative Works thereof in any medium, with or without
+   modifications, and in Source or Object form, provided that You
+   meet the following conditions:
+
+   (a) You must give any other recipients of the Work or
+       Derivative Works a copy of this License; and
+
+   (b) You must cause any modified files to carry prominent notices
+       stating that You changed the files; and
+
+   (c) You must retain, in the Source form of any Derivative Works
+       that You distribute, all copyright, patent, trademark, and
+       attribution notices from the Source form of the Work,
+       excluding those notices that do not pertain to any part of
+       the Derivative Works; and
+
+   (d) If the Work includes a "NOTICE" text file as part of its
+       distribution, then any Derivative Works that You distribute must
+       include a readable copy of the attribution notices contained
+       within such NOTICE file, excluding those notices that do not
+       pertain to any part of the Derivative Works, in at least one
+       of the following places: within a NOTICE text file distributed
+       as part of the Derivative Works; within the Source form or
+       documentation, if provided along with the Derivative Works; or,
+       within a display generated by the Derivative Works, if and
+       wherever such third-party notices normally appear. The contents
+       of the NOTICE file are for informational purposes only and
+       do not modify the License. You may add Your own attribution
+       notices within Derivative Works that You distribute, alongside
+       or as an addendum to the NOTICE text from the Work, provided
+       that such additional attribution notices cannot be construed
+       as modifying the License.
+
+   You may add Your own copyright statement to Your modifications and
+   may provide additional or different license terms and conditions
+   for use, reproduction, or distribution of Your modifications, or
+   for any such Derivative Works as a whole, provided Your use,
+   reproduction, and distribution of the Work otherwise complies with
+   the conditions stated in this License.
+
+5. Submission of Contributions. Unless You explicitly state otherwise,
+   any Contribution intentionally submitted for inclusion in the Work
+   by You to the Licensor shall be under the terms and conditions of
+   this License, without any additional terms or conditions.
+   Notwithstanding the above, nothing herein shall supersede or modify
+   the terms of any separate license agreement you may have executed
+   with Licensor regarding such Contributions.
+
+6. Trademarks. This License does not grant permission to use the trade
+   names, trademarks, service marks, or product names of the Licensor,
+   except as required for reasonable and customary use in describing the
+   origin of the Work and reproducing the content of the NOTICE file.
+
+7. Disclaimer of Warranty. Unless required by applicable law or
+   agreed to in writing, Licensor provides the Work (and each
+   Contributor provides its Contributions) on an "AS IS" BASIS,
+   WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or
+   implied, including, without limitation, any warranties or conditions
+   of TITLE, NON-INFRINGEMENT, MERCHANTABILITY, or FITNESS FOR A
+   PARTICULAR PURPOSE. You are solely responsible for determining the
+   appropriateness of using or redistributing the Work and assume any
+   risks associated with Your exercise of permissions under this License.
+
+8. Limitation of Liability. In no event and under no legal theory,
+   whether in tort (including negligence), contract, or otherwise,
+   unless required by applicable law (such as deliberate and grossly
+   negligent acts) or agreed to in writing, shall any Contributor be
+   liable to You for damages, including any direct, indirect, special,
+   incidental, or consequential damages of any character arising as a
+   result of this License or out of the use or inability to use the
+   Work (including but not limited to damages for loss of goodwill,
+   work stoppage, computer failure or malfunction, or any and all
+   other commercial damages or losses), even if such Contributor
+   has been advised of the possibility of such damages.
+
+9. Accepting Warranty or Additional Liability. While redistributing
+   the Work or Derivative Works thereof, You may choose to offer,
+   and charge a fee for, acceptance of support, warranty, indemnity,
+   or other liability obligations and/or rights consistent with this
+   License. However, in accepting such obligations, You may act only
+   on Your own behalf and on Your sole responsibility, not on behalf
+   of any other Contributor, and only if You agree to indemnify,
+   defend, and hold each Contributor harmless for any liability
+   incurred by, or claims asserted against, such Contributor by reason
+   of your accepting any such warranty or additional liability.
+
+END OF TERMS AND CONDITIONS
+
+Copyright 2026 Mohammad Movahedi
+
+Licensed under the Apache License, Version 2.0 (the "License");
+you may not use this file except in compliance with the License.
+You may obtain a copy of the License at
+
+    http://www.apache.org/licenses/LICENSE-2.0
+
+Unless required by applicable law or agreed to in writing, software
+distributed under the License is distributed on an "AS IS" BASIS,
+WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+See the License for the specific language governing permissions and
+limitations under the License.

strata_bench-1.1.0/NOTICE

@@ -0,0 +1,13 @@
+STRATA-Bench
+Copyright 2026 Mohammad Movahedi
+
+This product includes software developed for the STRATA-Bench evaluation
+protocol (Spatial-Temporal Retrieval, Alignment, Transparency & Audit).
+
+The Meridian Metro sandbox (MREB, UrbanPulse, LeaseWatch, district codes
+C-1 and C-8) is synthetic evaluation data. It is not affiliated with any
+real real-estate board, consultancy, or municipality.
+
+Live-track task text may name third-party publishers (e.g. TRREB, Urbanation)
+as research context. Those names remain the marks of their owners. STRATA-Bench
+does not redistribute their tables.

strata_bench-1.1.0/PKG-INFO

@@ -0,0 +1,218 @@
+Metadata-Version: 2.5
+Name: strata-bench
+Version: 1.1.0
+Summary: STRATA-Bench: Spatial-Temporal Retrieval, Alignment, Transparency & Audit — a benchmark for AI agents on fragmented geographic and longitudinal market intelligence.
+Project-URL: Homepage, https://github.com/movahedi-ca/strata-bench
+Project-URL: Documentation, https://github.com/movahedi-ca/strata-bench/tree/main/docs
+Project-URL: Issues, https://github.com/movahedi-ca/strata-bench/issues
+Project-URL: Source, https://github.com/movahedi-ca/strata-bench
+Author-email: Mohammad Movahedi <movahedi.m@northeastern.edu>
+License: Apache-2.0
+License-File: LICENSE
+License-File: NOTICE
+Keywords: agents,benchmark,evaluation,hallucination,llm,real-estate,retrieval,spatial-data
+Classifier: Development Status :: 4 - Beta
+Classifier: Intended Audience :: Science/Research
+Classifier: License :: OSI Approved :: Apache Software License
+Classifier: Programming Language :: Python :: 3
+Classifier: Programming Language :: Python :: 3.10
+Classifier: Programming Language :: Python :: 3.11
+Classifier: Programming Language :: Python :: 3.12
+Classifier: Topic :: Scientific/Engineering :: Artificial Intelligence
+Requires-Python: >=3.10
+Requires-Dist: pydantic>=2.6
+Provides-Extra: dev
+Requires-Dist: pytest>=8.0; extra == 'dev'
+Requires-Dist: ruff>=0.6; extra == 'dev'
+Description-Content-Type: text/markdown
+
+# STRATA-Bench
+
+![STRATA-Bench: layered glass strata of fragmented market data examined under a magnifying lens](docs/assets/hero.png)
+
+**Spatial-Temporal Retrieval, Alignment, Transparency & Audit**
+
+A benchmark for AI agents that must compile longitudinal market intelligence from **fragmented, geographically nested, and temporally incomplete** evidence — without hallucinating downtown numbers, silently interpolating missing quarters, or plotting gaps as if they were data.
+
+[![CI](https://github.com/movahedi-ca/strata-bench/actions/workflows/ci.yml/badge.svg)](https://github.com/movahedi-ca/strata-bench/actions/workflows/ci.yml)
+[![Coverage](https://codecov.io/gh/movahedi-ca/strata-bench/branch/main/graph/badge.svg)](https://codecov.io/gh/movahedi-ca/strata-bench)
+[![License](https://img.shields.io/badge/license-Apache%202.0-0e1210)](LICENSE)
+[![Python](https://img.shields.io/badge/python-3.10%2B-d4a574)](pyproject.toml)
+[![Version](https://img.shields.io/badge/version-1.1.0-6fbf9a)](CITATION.cff)
+
+📖 **Docs:** <https://movahedi-ca.github.io/strata-bench/> · 🗂️ **Tasks:** [`data/public/tasks.json`](data/public/tasks.json) · 📜 **Protocol:** [`docs/protocol.md`](docs/protocol.md)
+
+> **Headline finding from the motivating audit.** When six frontier engines were asked for a five-year downtown condominium rent series, they split into three archetypes: *Empirical Literalists* (faithful to the public board, honest about geography), *Synthetic Econometricians* (modeled a downtown premium and said so), and *Fragmented Extractors* (OCR'd charts, skipped nine of twenty quarters, and plotted those gaps as equal ticks). STRATA-Bench turns that failure mode into a scored protocol.
+
+```
+Headline score  =  0.25·Temporal  +  0.20·Source  +  0.15·Spatial
+                +  0.15·Statistical  +  0.15·Visual  +  0.10·Narrative
+Hard fails (geographic masking, silent interpolation, categorical axes on gappy
+series, OCR laundering, paywall fabrication) cap a task at 55% of earned credit.
+```
+
+---
+
+## Why this exists
+
+SWE-bench asks whether an agent can patch a repo. GAIA asks whether it can chain tools to answer a question. Neither asks whether the agent **knows what a number is**.
+
+In real institutional research — housing, labour, crime, wait times — the public file is almost never the geography the user named. The Toronto Regional Real Estate Board publishes a continuous GTA series; downtown districts C01/C08 are paywalled. Consultancy notes arrive on a different cadence, in $/sq ft. Listing-site charts drop quarters and then space the survivors evenly.
+
+Agents fail in *structured* ways:
+
+| Failure | What it looks like | STRATA family |
+| --- | --- | --- |
+| Geographic masking | Metro board figures labelled “Downtown Core” | **GEO** |
+| Temporal attrition | 11 of 20 quarters; 2026 trough silently dropped | **TEMP** |
+| OCR laundering | Rounded chart-readouts cited as MLS extracts | **SRC** |
+| Silent interpolation | Linear fill presented as reported | **INT** |
+| Non-uniform axis | 18-month hole drawn as wide as one quarter | **AXIS** |
+| Source blending | Urbanation psf + board CAD + listing-site guesses, one line | **SYN** |
+
+STRATA-Bench scores agents on whether they **disclose, tag, refuse, or model** — not on whether they emit a pretty dashboard.
+
+## Two tracks
+
+| Track | n | Reproducible? | Evidence |
+| --- | --- | --- | --- |
+| **Sandbox** (official) | 36 | Yes | Self-contained corpus: fictional **Meridian Metro** board (MREB), Core-district consultancy (UrbanPulse), gappy listing-site (LeaseWatch), plus adversarial trap files |
+| **Live** (optional probe) | 3 | No | Public-web Toronto / TRREB-style retrieval. Not used for the headline leaderboard |
+
+The sandbox cycle is *structurally* the 2021–2026 North American rental boom-bust (post-shock trough → 2023 peak → completion flood → 2026-Q1 trough → 2026-Q2 first rebound) with **held-out gold**. Numbers are not Toronto's. Geography codes are `metro` vs `core` (C-1 Harbour + C-8 Civic Core). Core gold is metro × **1.115**, never published as a public board table.
+
+## Task families
+
+| Family | n | Primary question |
+| --- | --- | --- |
+| `GEO` | 6 | Did the agent distinguish metro from core, or relabel? |
+| `TEMP` | 6 | Are all 20 quarters present, including the 2026 turning points? |
+| `SRC` | 6 | Is every cell sourced? Were visual estimates laundered? |
+| `INT` | 6 | Are interpolations tagged (`status=interpolated`, hollow markers)? |
+| `AXIS` | 6 | Is the x-axis datetime, with gaps as gaps? |
+| `SYN` | 6 | Can the agent narrate regimes without blending incompatible series? |
+| `LIVE` | 3 | Optional public-web analog |
+
+Full prompts: [`data/public/tasks.json`](data/public/tasks.json) · protocol: [`docs/protocol.md`](docs/protocol.md).
+
+## Install
+
+```bash
+git clone https://github.com/movahedi-ca/strata-bench.git
+cd strata-bench
+python -m pip install -e ".[dev]"
+python scripts/generate_sandbox.py   # corpus + gold + task catalog
+```
+
+Requires Python 3.10+.
+
+## Evaluate a submission
+
+Submissions are JSON documents matching [`submissions/SCHEMA.md`](submissions/SCHEMA.md).
+
+```bash
+strata-bench list-tasks
+strata-bench evaluate submissions/examples/literalist.json
+strata-bench evaluate submissions/examples/literalist.json --format markdown -o scorecard.md
+strata-bench evaluate path/to/run.json --format json
+```
+
+Python:
+
+```python
+from pathlib import Path
+from strata_bench import evaluate_submission
+
+result = evaluate_submission(Path("submissions/examples/literalist.json"))
+print(result["headline_score"], result["hard_fails"])
+```
+
+## Seed leaderboard (sandbox, v1.0 harness)
+
+These are *reference agents* shipped with the repo — not a claim about any vendor. They exist so the scorer has regression fixtures.
+
+| Agent | Archetype | Headline | Characteristic failure |
+| --- | --- | --- | --- |
+| `literalist` | Empirical Literalist | high | Refuses Core when unpaid; complete metro series |
+| `econometrician` | Synthetic Econometrician | high | Documents 11.5% premium; tags modeled cells |
+| `extractor` | Fragmented Extractor | low | OCR chart, categorical axis, truncated 2025, geographic mask |
+
+Run them:
+
+```bash
+strata-bench evaluate submissions/examples/literalist.json
+strata-bench evaluate submissions/examples/econometrician.json
+strata-bench evaluate submissions/examples/extractor.json
+```
+
+## Community leaderboard
+
+Ran your own agent? Add the run as `submissions/community/<agent>-<model>.json`
+(matching [`submissions/SCHEMA.md`](submissions/SCHEMA.md)) and open a pull
+request — the **Community leaderboard** workflow scores it against the frozen
+v1.0 gold and posts the scorecard as a comment on your PR.
+
+| Agent | Model | Headline | Run |
+| --- | --- | --- | --- |
+| *your agent here* | — | — | open a PR to claim this row |
+
+See [`submissions/community/README.md`](submissions/community/README.md) for the rules.
+
+## Protocol in one page
+
+1. **Name the geography of every number.** `metro` and `core` are different series. “Downtown” is not a board code.
+2. **Name the source of every number.** MREB public ≠ UrbanPulse ≠ LeaseWatch.
+3. **Do not invent paywalled district tables.** Disclose, model with a flagged premium, or refuse.
+4. **Missing ≠ zero.** Unpublished quarters are omitted or interpolated; they are never `status=reported`.
+5. **Plot time as time.** `axis_type=datetime`. Gaps occupy proportional width.
+6. **Do not blend units.** CAD rent and $/psf do not share a y-axis without conversion footnotes.
+7. **Mark estimates.** Hollow markers, asterisks with a legend, `interpolated_periods`.
+
+The unconstrained prompt *“give me downtown rents over five years”* is a trap. A protocol-compliant brief names the **provider**, **geography code**, **bedroom cut**, and **missing-quarter policy**. See [`docs/prompting.md`](docs/prompting.md).
+
+## Repository layout
+
+```
+strata-bench/
+├── strata_bench/          # scorer, schema, CLI
+├── data/
+│   ├── public/            # task catalog, source manifest (distributed)
+│   ├── hidden/gold.json   # held-out answers (used by the harness; do not give to agents)
+│   └── sandbox/corpus/    # the only legal evidence on Track Sandbox
+├── submissions/examples/  # three archetypal runs
+├── docs/                  # protocol, scoring, paper notes
+└── tests/
+```
+
+Agents on Track Sandbox may read `data/sandbox/corpus/**` and `data/public/tasks.json`. They may **not** read `data/hidden/`.
+
+## Citation
+
+```bibtex
+@software{movahedi2026strata,
+  author    = {Movahedi, Mohammad},
+  title     = {{STRATA-Bench}: Evaluating AI Agents on Fragmented Spatial-Temporal Market Intelligence},
+  year      = {2026},
+  version   = {1.1.0},
+  url       = {https://github.com/movahedi-ca/strata-bench},
+  license   = {Apache-2.0}
+}
+```
+
+If you use the Toronto *motivating audit* in related work, cite the protocol's ancestry: a 2026 comparative evaluation of six frontier engines on a five-year condominium rental retrospective, which isolated the geographic-masking trap, OCR graph reading, non-uniform temporal axes, and interpolation opacity as independent failure vectors.
+
+## Status
+
+**v1.1.0** — tooling & docs release. The benchmark itself (sandbox corpus,
+scorer, 36 official tasks) is **frozen at v1.0**. Live track is a probe, not a
+leaderboard. Gold is shipped in-tree so local evaluation is possible; a hashed
+hidden split for a future hosted leaderboard is sketched in
+`data/hidden/README.md` and tracked in
+[#1](https://github.com/movahedi-ca/strata-bench/issues/1).
+See [CHANGELOG.md](CHANGELOG.md).
+
+Contributions: [`CONTRIBUTING.md`](CONTRIBUTING.md). Code of conduct: [`CODE_OF_CONDUCT.md`](CODE_OF_CONDUCT.md). Security: [`SECURITY.md`](SECURITY.md).
+
+## License
+
+Apache-2.0. The **Meridian Metro** sandbox is synthetic evaluation data and is not a real board, city, or consultancy. Live-track citations of third-party research remain the property of those publishers; STRATA-Bench does not redistribute paywalled tables.

strata_bench-1.1.0/README.md

@@ -0,0 +1,190 @@
+# STRATA-Bench
+
+![STRATA-Bench: layered glass strata of fragmented market data examined under a magnifying lens](docs/assets/hero.png)
+
+**Spatial-Temporal Retrieval, Alignment, Transparency & Audit**
+
+A benchmark for AI agents that must compile longitudinal market intelligence from **fragmented, geographically nested, and temporally incomplete** evidence — without hallucinating downtown numbers, silently interpolating missing quarters, or plotting gaps as if they were data.
+
+[![CI](https://github.com/movahedi-ca/strata-bench/actions/workflows/ci.yml/badge.svg)](https://github.com/movahedi-ca/strata-bench/actions/workflows/ci.yml)
+[![Coverage](https://codecov.io/gh/movahedi-ca/strata-bench/branch/main/graph/badge.svg)](https://codecov.io/gh/movahedi-ca/strata-bench)
+[![License](https://img.shields.io/badge/license-Apache%202.0-0e1210)](LICENSE)
+[![Python](https://img.shields.io/badge/python-3.10%2B-d4a574)](pyproject.toml)
+[![Version](https://img.shields.io/badge/version-1.1.0-6fbf9a)](CITATION.cff)
+
+📖 **Docs:** <https://movahedi-ca.github.io/strata-bench/> · 🗂️ **Tasks:** [`data/public/tasks.json`](data/public/tasks.json) · 📜 **Protocol:** [`docs/protocol.md`](docs/protocol.md)
+
+> **Headline finding from the motivating audit.** When six frontier engines were asked for a five-year downtown condominium rent series, they split into three archetypes: *Empirical Literalists* (faithful to the public board, honest about geography), *Synthetic Econometricians* (modeled a downtown premium and said so), and *Fragmented Extractors* (OCR'd charts, skipped nine of twenty quarters, and plotted those gaps as equal ticks). STRATA-Bench turns that failure mode into a scored protocol.
+
+```
+Headline score  =  0.25·Temporal  +  0.20·Source  +  0.15·Spatial
+                +  0.15·Statistical  +  0.15·Visual  +  0.10·Narrative
+Hard fails (geographic masking, silent interpolation, categorical axes on gappy
+series, OCR laundering, paywall fabrication) cap a task at 55% of earned credit.
+```
+
+---
+
+## Why this exists
+
+SWE-bench asks whether an agent can patch a repo. GAIA asks whether it can chain tools to answer a question. Neither asks whether the agent **knows what a number is**.
+
+In real institutional research — housing, labour, crime, wait times — the public file is almost never the geography the user named. The Toronto Regional Real Estate Board publishes a continuous GTA series; downtown districts C01/C08 are paywalled. Consultancy notes arrive on a different cadence, in $/sq ft. Listing-site charts drop quarters and then space the survivors evenly.
+
+Agents fail in *structured* ways:
+
+| Failure | What it looks like | STRATA family |
+| --- | --- | --- |
+| Geographic masking | Metro board figures labelled “Downtown Core” | **GEO** |
+| Temporal attrition | 11 of 20 quarters; 2026 trough silently dropped | **TEMP** |
+| OCR laundering | Rounded chart-readouts cited as MLS extracts | **SRC** |
+| Silent interpolation | Linear fill presented as reported | **INT** |
+| Non-uniform axis | 18-month hole drawn as wide as one quarter | **AXIS** |
+| Source blending | Urbanation psf + board CAD + listing-site guesses, one line | **SYN** |
+
+STRATA-Bench scores agents on whether they **disclose, tag, refuse, or model** — not on whether they emit a pretty dashboard.
+
+## Two tracks
+
+| Track | n | Reproducible? | Evidence |
+| --- | --- | --- | --- |
+| **Sandbox** (official) | 36 | Yes | Self-contained corpus: fictional **Meridian Metro** board (MREB), Core-district consultancy (UrbanPulse), gappy listing-site (LeaseWatch), plus adversarial trap files |
+| **Live** (optional probe) | 3 | No | Public-web Toronto / TRREB-style retrieval. Not used for the headline leaderboard |
+
+The sandbox cycle is *structurally* the 2021–2026 North American rental boom-bust (post-shock trough → 2023 peak → completion flood → 2026-Q1 trough → 2026-Q2 first rebound) with **held-out gold**. Numbers are not Toronto's. Geography codes are `metro` vs `core` (C-1 Harbour + C-8 Civic Core). Core gold is metro × **1.115**, never published as a public board table.
+
+## Task families
+
+| Family | n | Primary question |
+| --- | --- | --- |
+| `GEO` | 6 | Did the agent distinguish metro from core, or relabel? |
+| `TEMP` | 6 | Are all 20 quarters present, including the 2026 turning points? |
+| `SRC` | 6 | Is every cell sourced? Were visual estimates laundered? |
+| `INT` | 6 | Are interpolations tagged (`status=interpolated`, hollow markers)? |
+| `AXIS` | 6 | Is the x-axis datetime, with gaps as gaps? |
+| `SYN` | 6 | Can the agent narrate regimes without blending incompatible series? |
+| `LIVE` | 3 | Optional public-web analog |
+
+Full prompts: [`data/public/tasks.json`](data/public/tasks.json) · protocol: [`docs/protocol.md`](docs/protocol.md).
+
+## Install
+
+```bash
+git clone https://github.com/movahedi-ca/strata-bench.git
+cd strata-bench
+python -m pip install -e ".[dev]"
+python scripts/generate_sandbox.py   # corpus + gold + task catalog
+```
+
+Requires Python 3.10+.
+
+## Evaluate a submission
+
+Submissions are JSON documents matching [`submissions/SCHEMA.md`](submissions/SCHEMA.md).
+
+```bash
+strata-bench list-tasks
+strata-bench evaluate submissions/examples/literalist.json
+strata-bench evaluate submissions/examples/literalist.json --format markdown -o scorecard.md
+strata-bench evaluate path/to/run.json --format json
+```
+
+Python:
+
+```python
+from pathlib import Path
+from strata_bench import evaluate_submission
+
+result = evaluate_submission(Path("submissions/examples/literalist.json"))
+print(result["headline_score"], result["hard_fails"])
+```
+
+## Seed leaderboard (sandbox, v1.0 harness)
+
+These are *reference agents* shipped with the repo — not a claim about any vendor. They exist so the scorer has regression fixtures.
+
+| Agent | Archetype | Headline | Characteristic failure |
+| --- | --- | --- | --- |
+| `literalist` | Empirical Literalist | high | Refuses Core when unpaid; complete metro series |
+| `econometrician` | Synthetic Econometrician | high | Documents 11.5% premium; tags modeled cells |
+| `extractor` | Fragmented Extractor | low | OCR chart, categorical axis, truncated 2025, geographic mask |
+
+Run them:
+
+```bash
+strata-bench evaluate submissions/examples/literalist.json
+strata-bench evaluate submissions/examples/econometrician.json
+strata-bench evaluate submissions/examples/extractor.json
+```
+
+## Community leaderboard
+
+Ran your own agent? Add the run as `submissions/community/<agent>-<model>.json`
+(matching [`submissions/SCHEMA.md`](submissions/SCHEMA.md)) and open a pull
+request — the **Community leaderboard** workflow scores it against the frozen
+v1.0 gold and posts the scorecard as a comment on your PR.
+
+| Agent | Model | Headline | Run |
+| --- | --- | --- | --- |
+| *your agent here* | — | — | open a PR to claim this row |
+
+See [`submissions/community/README.md`](submissions/community/README.md) for the rules.
+
+## Protocol in one page
+
+1. **Name the geography of every number.** `metro` and `core` are different series. “Downtown” is not a board code.
+2. **Name the source of every number.** MREB public ≠ UrbanPulse ≠ LeaseWatch.
+3. **Do not invent paywalled district tables.** Disclose, model with a flagged premium, or refuse.
+4. **Missing ≠ zero.** Unpublished quarters are omitted or interpolated; they are never `status=reported`.
+5. **Plot time as time.** `axis_type=datetime`. Gaps occupy proportional width.
+6. **Do not blend units.** CAD rent and $/psf do not share a y-axis without conversion footnotes.
+7. **Mark estimates.** Hollow markers, asterisks with a legend, `interpolated_periods`.
+
+The unconstrained prompt *“give me downtown rents over five years”* is a trap. A protocol-compliant brief names the **provider**, **geography code**, **bedroom cut**, and **missing-quarter policy**. See [`docs/prompting.md`](docs/prompting.md).
+
+## Repository layout
+
+```
+strata-bench/
+├── strata_bench/          # scorer, schema, CLI
+├── data/
+│   ├── public/            # task catalog, source manifest (distributed)
+│   ├── hidden/gold.json   # held-out answers (used by the harness; do not give to agents)
+│   └── sandbox/corpus/    # the only legal evidence on Track Sandbox
+├── submissions/examples/  # three archetypal runs
+├── docs/                  # protocol, scoring, paper notes
+└── tests/
+```
+
+Agents on Track Sandbox may read `data/sandbox/corpus/**` and `data/public/tasks.json`. They may **not** read `data/hidden/`.
+
+## Citation
+
+```bibtex
+@software{movahedi2026strata,
+  author    = {Movahedi, Mohammad},
+  title     = {{STRATA-Bench}: Evaluating AI Agents on Fragmented Spatial-Temporal Market Intelligence},
+  year      = {2026},
+  version   = {1.1.0},
+  url       = {https://github.com/movahedi-ca/strata-bench},
+  license   = {Apache-2.0}
+}
+```
+
+If you use the Toronto *motivating audit* in related work, cite the protocol's ancestry: a 2026 comparative evaluation of six frontier engines on a five-year condominium rental retrospective, which isolated the geographic-masking trap, OCR graph reading, non-uniform temporal axes, and interpolation opacity as independent failure vectors.
+
+## Status
+
+**v1.1.0** — tooling & docs release. The benchmark itself (sandbox corpus,
+scorer, 36 official tasks) is **frozen at v1.0**. Live track is a probe, not a
+leaderboard. Gold is shipped in-tree so local evaluation is possible; a hashed
+hidden split for a future hosted leaderboard is sketched in
+`data/hidden/README.md` and tracked in
+[#1](https://github.com/movahedi-ca/strata-bench/issues/1).
+See [CHANGELOG.md](CHANGELOG.md).
+
+Contributions: [`CONTRIBUTING.md`](CONTRIBUTING.md). Code of conduct: [`CODE_OF_CONDUCT.md`](CODE_OF_CONDUCT.md). Security: [`SECURITY.md`](SECURITY.md).
+
+## License
+
+Apache-2.0. The **Meridian Metro** sandbox is synthetic evaluation data and is not a real board, city, or consultancy. Live-track citations of third-party research remain the property of those publishers; STRATA-Bench does not redistribute paywalled tables.