Metadata-Version: 2.4
Name: concerto-multirobot
Version: 0.8.2
Summary: Heterogeneous Multi-Robot Ad-Hoc Teamwork — benchmark + safety stack.
Project-URL: Homepage, https://github.com/fsafaei/concerto
Project-URL: Documentation, https://fsafaei.github.io/concerto/
Project-URL: Repository, https://github.com/fsafaei/concerto
Project-URL: Issues, https://github.com/fsafaei/concerto/issues
Project-URL: Changelog, https://github.com/fsafaei/concerto/blob/main/CHANGELOG.md
Author-email: Farhad Safaei <cooperative.physical.ai@gmail.com>
License: 
                                         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
        
           APPENDIX: How to apply the Apache License to your work.
        
              To apply the Apache License to your work, attach the following
              boilerplate notice, with the fields enclosed by brackets "[]"
              replaced with your own identifying information. (Don't include
              the brackets!)  The text should be enclosed in the appropriate
              comment syntax for the file format. We also recommend that a
              file or class name and description of purpose be included on the
              same "printed page" as the copyright notice for easier
              identification within third-party archives.
        
           Copyright 2026 Farhad Safaei
        
           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.
License-File: LICENSE
License-File: NOTICE
Keywords: ad-hoc-teamwork,conformal-prediction,control-barrier-functions,heterogeneous-marl,manipulation,multi-robot,robotics-benchmark
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 :: 3.11
Classifier: Programming Language :: Python :: 3.12
Classifier: Topic :: Scientific/Engineering :: Artificial Intelligence
Requires-Python: <3.13,>=3.11
Requires-Dist: cvxpy<2.0,>=1.8.2
Requires-Dist: gymnasium<2.0,>=1.3.0
Requires-Dist: harl-aht<0.2.0,>=0.1.0
Requires-Dist: hydra-core<2.0,>=1.3.2
Requires-Dist: mani-skill==3.0.1
Requires-Dist: numpy<2.5,>=1.26
Requires-Dist: omegaconf<3.0,>=2.3
Requires-Dist: pydantic<3.0,>=2.6
Requires-Dist: pytorch-kinematics<1.0,>=0.7
Requires-Dist: qpsolvers[clarabel,osqp]>=4.0
Requires-Dist: rich>=15.0.0
Requires-Dist: sapien<4.0,>=3.0
Requires-Dist: scipy<2.0,>=1.17.1
Requires-Dist: structlog>=24.1
Requires-Dist: torch<3.0,>=2.9
Requires-Dist: wandb<1.0,>=0.17
Requires-Dist: warp-lang<2.0,>=1.13.0
Provides-Extra: dev
Requires-Dist: commitizen>=3.27; extra == 'dev'
Requires-Dist: cyclonedx-bom>=4.0; extra == 'dev'
Requires-Dist: hypothesis>=6.108; extra == 'dev'
Requires-Dist: ipykernel; extra == 'dev'
Requires-Dist: ipython; extra == 'dev'
Requires-Dist: pip-licenses>=4.0; extra == 'dev'
Requires-Dist: pre-commit>=3.7; extra == 'dev'
Requires-Dist: pyright>=1.1.409; extra == 'dev'
Requires-Dist: pytest-benchmark>=4.0; extra == 'dev'
Requires-Dist: pytest-cov>=5.0; extra == 'dev'
Requires-Dist: pytest-xdist>=3.6; extra == 'dev'
Requires-Dist: pytest>=8.2; extra == 'dev'
Requires-Dist: ruff>=0.6; extra == 'dev'
Provides-Extra: docs
Requires-Dist: mike>=2.0; extra == 'docs'
Requires-Dist: mkdocs-material>=9.5; extra == 'docs'
Requires-Dist: mkdocs>=1.6.1; extra == 'docs'
Requires-Dist: mkdocstrings[python]>=0.25; extra == 'docs'
Provides-Extra: eval
Requires-Dist: pandas<3.0,>=1.4; extra == 'eval'
Requires-Dist: rliable<2.0,>=1.0; extra == 'eval'
Provides-Extra: observability
Requires-Dist: imageio[ffmpeg]>=2.31; extra == 'observability'
Provides-Extra: viz
Requires-Dist: matplotlib>=3.9; extra == 'viz'
Requires-Dist: seaborn>=0.13.2; extra == 'viz'
Description-Content-Type: text/markdown

<p align="center">
  <a href="https://github.com/fsafaei/concerto">
    <picture>
      <source media="(prefers-color-scheme: dark)" srcset="docs/assets/dark_mode.png">
      <img alt="CONCERTO" src="docs/assets/light_mode.png" width="420"/>
    </picture>
  </a>
</p>

<p align="center">
  <em>Contact-rich coordination with opaque, heterogeneous teammates.</em><br/>
  <strong>CONCERTO</strong> is the method.
  <strong>CHAMBER</strong> is the benchmark.
  We evaluate CONCERTO on CHAMBER.
</p>

<p align="center">
  <a href="https://doi.org/10.5281/zenodo.20128468">
    <img alt="DOI" src="https://zenodo.org/badge/DOI/10.5281/zenodo.20128468.svg"/></a>
  <a href="https://github.com/fsafaei/concerto/actions/workflows/ci.yml">
    <img alt="CI"   src="https://github.com/fsafaei/concerto/actions/workflows/ci.yml/badge.svg"/></a>
  <a href="https://fsafaei.github.io/concerto/">
    <img alt="Docs" src="https://img.shields.io/badge/docs-latest-blue"/></a>
  <a href="LICENSE">
    <img alt="License" src="https://img.shields.io/badge/license-Apache%202.0-green"/></a>
  <a href="https://scorecard.dev/viewer/?uri=github.com/fsafaei/concerto">
    <img alt="OpenSSF Scorecard" src="https://api.scorecard.dev/projects/github.com/fsafaei/concerto/badge"/></a>
  <a href="https://test.pypi.org/project/concerto-multirobot/">
    <img alt="TestPyPI" src="https://img.shields.io/pypi/v/concerto-multirobot?pypiBaseUrl=https%3A%2F%2Ftest.pypi.org&amp;label=TestPyPI"/></a>
</p>

## What this is

CHAMBER is a simulation benchmark for **physically-coupled ad-hoc
cooperation**: two robot arms jointly manipulate one object (carry a
rigid bar; hand a part over and place it) while the evaluated robot —
the *ego* — works with a **black-box partner**, meaning a teammate
whose policy, parameters, and training history it cannot see and that
it was never trained with. CONCERTO is the companion method: a
safety-filter stack and a training loop for exactly this setting.
This repository ships both, plus everything needed to reproduce and
verify every number below.

**The claim this benchmark supports.** A leaderboard row supports one
kind of claim: *this method, controlling one arm, achieves the stated
joint-success rate with previously-unseen black-box partners on this
physically-coupled task, with the stated confidence interval.* Every
row is recomputed at render time from a committed result bundle that
anyone can verify with one command. Nothing broader is claimed — no
cross-task ranking, no scalar composite, and no
heterogeneity-robustness aggregate in v1.0.

**Under what assumptions.** Both scored tasks passed a preregistered
admission protocol showing that a reference pair solves them, one
robot alone cannot, and destroying the physical coupling destroys
performance ([ADR-027](adr/ADR-027-chamber-bench-v1-protocol.md)).
Partners are frozen during evaluation and drawn from a versioned
partner zoo with a public/private split, so a method cannot overfit
the full evaluation set. CPU evaluation is byte-identical under the
pinned `uv.lock` and a root seed, which is what makes committed
bundles recomputable.

**With what limitations.** Everything is simulation; no real-robot
results exist. Teams are exactly two agents on the listed embodiments,
and the v1.0 partner zoo contains no vision-based partners. Measured
heterogeneity effects on co-carry are null so far, so v1.0 measures
ad-hoc-teamwork competence, not heterogeneity robustness — see
[Limitations](#limitations).

## Install

[`uv`](https://docs.astral.sh/uv/) is the package manager; versions
are pinned by the committed `uv.lock`. Do not invoke pip directly.

```bash
git clone https://github.com/fsafaei/concerto.git
cd concerto
pip install uv && uv sync --group dev
```

Python 3.11 or 3.12. The training stack (the HARL fork, published as
the `harl-aht` distribution) resolves automatically as a runtime
dependency; no extra install step is needed.

## Five-minute quickstart

Run one evaluation and verify one shipped result. The commands below
are executed verbatim by CI on every push (jobs `smoke-eval` and
`readme-quickstart` in
[`.github/workflows/ci.yml`](.github/workflows/ci.yml)), so they are
guaranteed to work on a clean checkout.

**1. Run a real evaluation on the CPU diagnostic task** (~1 minute).
This produces a *result bundle*: a directory containing every episode
record, the exact command, the git state, and SHA-256 checksums of all
files.

```bash
uv run chamber-eval run --task mpe_cooperative_push --policy random \
  --partner scripted_heuristic --seeds 2 --episodes 5 --out out/quickstart-bundle
uv run chamber-eval verify out/quickstart-bundle
```

Expected: the run reports `10 episodes; success IQM 0.667 [0.000, 1.000]`
(byte-identical on CPU under the pinned lockfile), and `verify` prints
a check table ending in `verify: PASS`. On a locally-modified checkout
add `--allow-dirty` to both commands; such bundles are marked dirty
and are ineligible for the leaderboard.

**2. Verify a shipped leaderboard row** (~10 seconds). This re-checks
every file hash in the committed bundle, recomputes the headline
statistics from the raw per-episode records, and re-derives the
preregistration tag:

```bash
uv run chamber-eval verify spikes/results/benchmark/cocarry-v1/ref-script-2026-07-05
```

Every cell of the leaderboard below can be checked the same way, and
[docs/how-to/reproduce-results.md](docs/how-to/reproduce-results.md)
gives the full re-run command for every row.

## The benchmark

The suite is pinned by the `chamber.tasks` registry
([ADR-027](adr/ADR-027-chamber-bench-v1-protocol.md) §Decision):
every task carries exactly one tier and an admission status backed by
committed evidence. `ADMITTED` is earned, never asserted —
`chamber-eval admission` executes the A1/A2/A3 protocol under a
tag-locked preregistration and writes the immutable report the status
cites (see `spikes/results/admission/`). The table is generated; CI
fails if it drifts.

<!-- CHAMBER-BENCH-TASK-TABLE:BEGIN — generated by scripts/render_task_table.py; edit src/chamber/tasks/, not this table -->

| Task | Tier | Status | Axis validity | Task card |
|------|------|--------|---------------|-----------|
| `mpe_cooperative_push@v1` — MPE cooperative push (CPU diagnostic) | 0 (rig diagnostics) | DIAGNOSTIC | all untested | [mpe_cooperative_push](docs/reference/tasks/mpe_cooperative_push.md) |
| `stage0_smoke@v1` — Stage-0 tri-embodiment smoke rig | 0 (rig diagnostics) | DIAGNOSTIC | all untested | [stage0_smoke](docs/reference/tasks/stage0_smoke.md) |
| `stage1_pickplace_as@v1` — Stage-1 pick-and-place — action-space (AS) control | 1 (controls) | CONTROL | AS `invalid` | [stage1_pickplace_as](docs/reference/tasks/stage1_pickplace_as.md) |
| `stage1_pickplace_om@v1` — Stage-1 pick-and-place — observation-modality (OM) control | 1 (controls) | CONTROL | all untested | [stage1_pickplace_om](docs/reference/tasks/stage1_pickplace_om.md) |
| `cocarry@v1` — Co-carry — rigid dual-arm bar transport | 2 (admitted cooperation tasks) | ADMITTED | AS `null` | [cocarry](docs/reference/tasks/cocarry.md) |
| `handover_place@v1` — Handover-and-place under takt pressure | 2 (admitted cooperation tasks) | ADMITTED | all untested | [handover_place](docs/reference/tasks/handover_place.md) |
| `amr_handover_dynamic@v0` — AMR dynamic handover (spec-only candidate) | 3 (documented candidates and closures) | CANDIDATE | all untested | [amr_handover_dynamic](docs/reference/tasks/amr_handover_dynamic.md) |
| `co_hold_secure@v0` — Co-hold-and-secure (spec-only candidate) | 3 (documented candidates and closures) | CANDIDATE | all untested | [co_hold_secure](docs/reference/tasks/co_hold_secure.md) |
| `coinsert@v1` — Co-insert — hold-and-insert (closed; open challenge) | 3 (documented candidates and closures) | CLOSED | all untested | [coinsert](docs/reference/tasks/coinsert.md) |

<!-- CHAMBER-BENCH-TASK-TABLE:END -->

## Leaderboard — CHAMBER-Bench v1.0

Per-task tables only; no scalar composite
([ADR-027](adr/ADR-027-chamber-bench-v1-protocol.md) §Decision,
reporting rules). Every row is recomputed from committed result
bundles that pass the full `chamber-eval verify` check table at render
time — an unverifiable entry refuses the whole render — under the
tag-locked campaign preregistration. Two rows carry honest labels and
are not competitors: REF-SCRIPT is the *oracle reference* (a scripted
controller with privileged access, shown as the solvability ceiling),
and B-JOINT is the *non-AHT upper anchor* (a jointly-trained pair
evaluated with its own training partner, which breaks the unseen-partner
condition by construction). The block is generated by
`scripts/render_leaderboard_table.py`; CI fails if it drifts.

<!-- CHAMBER-BENCH-LEADERBOARD:BEGIN — generated by scripts/render_leaderboard_table.py; edit spikes/results/benchmark/, not this table -->
#### cocarry-v1

| Row | Label | Success IQM [95% CI] | Success mean | Stress p90 (N) | Per-partner range | Seeds | Bundles |
|---|---|---|---|---|---|---|---|
| REF-SCRIPT | *oracle reference* | 1.000 [1.000, 1.000] | 1.000 | 102.0 | 1.00 (imp_stiff_low) … 1.00 (imp_stiff_low) | 0, 1, 2, 3, 4 | `spikes/results/benchmark/cocarry-v1/ref-script-2026-07-05` |
| B-RND |  | 0.000 [0.000, 0.000] | 0.001 | 277.8 | 0.00 (imp_stiff_low) … 0.00 (imp_blend_b) | 0, 1, 2, 3, 4 | `spikes/results/benchmark/cocarry-v1/b-rnd-2026-07-05` |
| B-STAT |  | 0.492 [0.396, 0.591] | 0.496 | 101.4 | 0.16 (imp_lag_bounded) … 0.66 (imp_blend_b) | 0, 1, 2, 3, 4 | `spikes/results/benchmark/cocarry-v1/b-stat-2026-07-05` |
| B-BLIND |  | 1.000 [1.000, 1.000] | 0.967 | 107.6 | 0.86 (imp_lag_bounded) … 1.00 (imp_stiff_low) | 0, 1, 2, 3, 4 | `spikes/results/benchmark/cocarry-v1/b-blind-2026-07-06` |
| B-AHT |  | 1.000 [1.000, 1.000] | 0.961 | 108.4 | 0.88 (imp_lag_bounded) … 1.00 (imp_stiff_low) | 0, 1, 2, 3, 4 | `spikes/results/benchmark/cocarry-v1/b-aht-2026-07-06` |
| B-JOINT | *non-AHT upper anchor* | 1.000 [1.000, 1.000] | 0.944 | 41.2 | 0.94 (frozen_cocarry_joint) … 0.94 (frozen_cocarry_joint) | 0, 1, 2, 3, 4 | `spikes/results/benchmark/cocarry-v1/b-joint-seed0-2026-07-06`<br>`spikes/results/benchmark/cocarry-v1/b-joint-seed1-2026-07-06`<br>`spikes/results/benchmark/cocarry-v1/b-joint-seed2-2026-07-06`<br>`spikes/results/benchmark/cocarry-v1/b-joint-seed3-2026-07-06`<br>`spikes/results/benchmark/cocarry-v1/b-joint-seed4-2026-07-06` |

*Solvability anchor: matched scripted reference 1.000 (admission A1, `spikes/results/admission/cocarry-2026-07-05/`).*
*Success IQM is the preregistered `iqm_success_rate`: the interquartile mean over per-episode success indicators; it is intentionally conservative away from saturation (e.g., handover REF-SCRIPT mean 0.338 vs IQM 0.176) and trims minority cells at saturation (e.g., B-AHT mean 0.961 vs IQM 1.000 — see the per-partner range).*

#### handover-v1

| Row | Label | Success IQM [95% CI] | Success mean | Stress p90 (N) | Per-partner range | Seeds | Bundles |
|---|---|---|---|---|---|---|---|
| REF-SCRIPT | *oracle reference* | 0.176 [0.072, 0.284] | 0.338 | 211.5 | 0.13 (presenter_mismatch_45) … 0.55 (presenter_mismatch_30) | 0, 1, 2, 3, 4 | `spikes/results/benchmark/handover-v1/ref-script-2026-07-06` |
| B-RND |  | 0.000 [0.000, 0.000] | 0.000 | 3858.1 | 0.00 (presenter_mismatch_30) … 0.00 (presenter_mismatch_30) | 0, 1, 2, 3, 4 | `spikes/results/benchmark/handover-v1/b-rnd-2026-07-06` |
| B-STAT |  | 0.048 [0.000, 0.140] | 0.274 | 225.6 | 0.05 (presenter_mismatch_45) … 0.50 (presenter_mismatch_30) | 0, 1, 2, 3, 4 | `spikes/results/benchmark/handover-v1/b-stat-2026-07-06` |

*Solvability anchor: matched presenter 1.000 (Gate-0 Limb 1, `spikes/results/handover-place-gate0-2026-06-26/`). Leaderboard cells are the measured coupling-valid (mismatch) region by design — the low oracle number is the mismatch penalty, not unsolvability.*
*Success IQM is the preregistered `iqm_success_rate`: the interquartile mean over per-episode success indicators; it is intentionally conservative away from saturation (e.g., handover REF-SCRIPT mean 0.338 vs IQM 0.176) and trims minority cells at saturation (e.g., B-AHT mean 0.961 vs IQM 1.000 — see the per-partner range).*
<!-- CHAMBER-BENCH-LEADERBOARD:END -->

Row ids follow the preregistered baseline set: B-RND (random
actions), B-STAT (static, hold-position), B-BLIND (learned ego with
partner observations masked), B-AHT (learned ad-hoc-teamwork ego).
Reproduction command for every row:
[docs/how-to/reproduce-results.md](docs/how-to/reproduce-results.md).

## Validity

A score on this benchmark means something only if the task actually
forces cooperation, so every claim rests on the coupling-validity
criterion of
[ADR-026](adr/ADR-026-coupling-validity-criterion.md). Tier-1 control
tasks are kept precisely because they are solvable by one robot alone:
they separate plain manipulation skill from cooperation, and a method
that gains on a control is measuring something other than teamwork.
Admission ([ADR-027](adr/ADR-027-chamber-bench-v1-protocol.md)) is
what a task must earn before its scores count, by three preregistered
checks: a reference pair solves it (A1), the best single robot cannot
(A2), and a scripted ego stripped of exactly its coupling channel
collapses (A3). Co-carry and handover-and-place passed all three;
pick-and-place passed A1 but failed A2 and A3 and is retained as a
Tier-1 control — the protocol demonstrably rejects tasks. The
heterogeneity-robustness score (HRS, ADR-008), the project's original
scalar aggregate, is suspended in v1.0 because no task×axis cell has
a validated heterogeneity effect yet, and averaging nulls into a
headline number would manufacture a ranking out of nothing. Until a
validated cell exists, per-task success tables with confidence
intervals are the only reported currency. The full protocol, with the
co-carry admission as a worked example, is at
[docs/explanation/evaluation-protocol.md](docs/explanation/evaluation-protocol.md).

## Submitting a leaderboard entry

Fork, run the preregistered cells against the public partner set,
open a PR containing your result bundle; CI re-runs
`chamber-eval verify` on it, and the maintainer may spot-check your
method against the private partner split before merge. Full protocol:
[docs/how-to/submit-leaderboard.md](docs/how-to/submit-leaderboard.md).

## Limitations

- **Simulation only.** All results are ManiSkill v3 / SAPIEN (plus
  one CPU diagnostic task); no real-robot results exist.
- **Two agents.** Every task is one ego and one partner; larger teams
  are out of scope for v1.0.
- **Listed embodiments only.** Results cover the embodiments named in
  the task cards (Panda-class arms on the scored rows); an xArm6
  variant was studied and is documented in ADR-026's revision log,
  but no leaderboard row uses it.
- **Heterogeneity effects are null on co-carry so far.** Partner-style
  and embodiment heterogeneity were measured and found null under
  fair matching (ADR-026 §Open questions); v1.0 therefore measures
  ad-hoc-teamwork competence with black-box partners, not
  heterogeneity robustness, and the HRS aggregate stays suspended.
- **No vision-based partners.** The v1.0 partner zoo is scripted
  impedance controllers, scripted presenters, and frozen learned
  policies; vision-language-action partners are deferred stubs.

## Documentation

Full documentation: [**fsafaei.github.io/concerto**](https://fsafaei.github.io/concerto/)

- [Evaluation protocol](docs/explanation/evaluation-protocol.md) — run classes, admission, reporting rules, versioning.
- [Reproduce the results](docs/how-to/reproduce-results.md) — one command block per leaderboard row.
- [Submit a leaderboard entry](docs/how-to/submit-leaderboard.md).
- [Datasheet](docs/reference/datasheet.md) — the released data artifacts, described.
- [Architecture](docs/explanation/architecture.md) — the two packages, the safety stack, repository layout.
- [Positioning](docs/explanation/positioning.md) — why this exists, prior work, the six heterogeneity axes, roadmap.
- [FAQ](docs/explanation/faq.md) · [Glossary](docs/reference/glossary.md) · [ADR index](adr/ADR-INDEX.md)
- [MAINTENANCE.md](MAINTENANCE.md) — who maintains this, versioning and deprecation discipline, contact.

## Contributing

Open from the first commit; PRs welcome. Read
[`CONTRIBUTING.md`](CONTRIBUTING.md) for the development flow,
commit-signing and DCO requirements, and the contributor license
agreement (ADR-012). Every PR cites the ADR section it touches;
protocol changes require a new ADR, never an edit to an accepted one.
Code of conduct: [`CODE_OF_CONDUCT.md`](CODE_OF_CONDUCT.md). Security
policy: [`SECURITY.md`](SECURITY.md).

## Citation

If you use CONCERTO or CHAMBER in your research, please cite the
archived software release:

```bibtex
@software{safaei2026concerto,
  author       = {Safaei, Farhad},
  title        = {{CONCERTO} and {CHAMBER}: Contact-rich Coordination
                  with Opaque, Heterogeneous Teammates},
  year         = {2026},
  publisher    = {Zenodo},
  doi          = {10.5281/zenodo.20128468},
  url          = {https://doi.org/10.5281/zenodo.20128468},
  note         = {arXiv preprint forthcoming},
}
```

Citation metadata is also in [`CITATION.cff`](CITATION.cff), so GitHub
renders a "Cite this repository" button.

## License

Apache 2.0. See [`LICENSE`](LICENSE) and [`NOTICE`](NOTICE). The
safety stack composes
[Wang et al. 2017](https://ieeexplore.ieee.org/document/7989121)
(decentralised exponential control barrier functions),
[Huriot & Sibai 2025](https://arxiv.org/abs/2409.18862) (conformal
control barrier functions), and
[Morton & Pavone 2025](https://arxiv.org/abs/2503.17678)
(operational-space control barrier functions). CHAMBER is a wrapper
layer over [ManiSkill v3](https://github.com/haosulab/ManiSkill) and
depends on a fork of [HARL](https://github.com/PKU-MARL/HARL) for the
training stack.
