SII-EricSanchez commited on
Commit
f78110c
·
verified ·
1 Parent(s): aedc502

docs: disable misleading imagefolder viewer and clarify code/data boundary

Browse files
Files changed (1) hide show
  1. README.md +78 -194
README.md CHANGED
@@ -1,5 +1,6 @@
1
  ---
2
- pretty_name: Act2Sim v0.7.0 public task data
 
3
  license: other
4
  license_name: mixed-upstream-licenses
5
  license_link: https://huggingface.co/datasets/SII-EricSanchez/act2sim-data/blob/v0.7.0/README.md#licenses-and-provenance
@@ -12,53 +13,56 @@ tags:
12
  - robotics
13
  ---
14
 
15
- # Act2Sim v0.7.0 public task data
16
 
17
- This revision changes drawer scoring and the peg task. Cloth is unchanged from `v0.6.0`.
18
- - **Drawer** uses a 72-probe panel and scores trajectories only: F has no event term. Both changes passed a sealed validity check against an independent behavior measure, where `v0.6.0` drawer scoring failed it.
19
- - **Peg** is task revision `peg-v2`:
20
- - The robot's fingers now read the peg's own friction and grip it with a stiff coupling.
21
- - A new 60-probe panel loosens the grip so that the peg slips.
22
- - F scores trajectories only and now responds to both the peg's mass and its friction.
23
- - It passed the same sealed validity check twice; `peg-v1` had failed it.
24
- - **Cloth** ships its interaction assets only. This revision publishes no calibrated cloth evaluation.
25
 
26
- Drawer and peg scores from `v0.6.0` and `v0.7.0` are not comparable. Compare results only within one task and protocol identity.
27
 
28
- Act2Sim evaluates worlds built through code:
29
- 1. An agent interacts with a target world.
30
- 2. It writes a source submission describing its reconstructed world.
31
- 3. It evaluates that submission on the user's own machine.
32
 
33
- The Python package supplies the API, simulation and evaluator. This dataset supplies the target assets, public robot assets, frozen evaluation data and perceptual metric weights. No hosted evaluator, agent upload, model-service account or benchmark-server compute is required.
34
 
35
- Public target and evaluation files are downloadable. The level-specific API controls the information supplied to the agent; it is not a security sandbox against a user reading downloaded evaluator files. Researchers are responsible for keeping their reconstruction agent within the declared information boundary, and for disclosing any use of additional target information.
 
 
 
36
 
37
- ## Tasks and levels
38
 
39
- | Task | Task revision | World | Robot configuration | Evaluation in this revision | Scoring device |
40
- | --- | --- | --- | --- | --- | --- |
41
- | `drawer/open_stop` | `drawer-v1` | Passive RoboCasa drawer opening and stopping | One FR3 workcell arm | Levels 1–3 | CPU |
42
- | `assembly/peg_insert` | `peg-v2` | Factory peg and hole; the probes grasp, loosen and shake the peg | One FR3 workcell arm | Levels 1–3 | Linux CUDA, logical `cuda:0` |
43
- | `cloth/unfold_shirt` | `shirt-v2` | Coarse DexGarmentLab shirt manipulation | Two FR3 workcell arms | Interaction only | — |
44
 
45
- | Level | Information provided through the agent API | Reported scores |
46
- | --- | --- | --- |
47
- | 1 | RGB, depth and robot state, plus initial exterior geometry and noisy whole-object poses | Physical F |
48
- | 2 | RGB, depth and robot state; initial target geometry and object poses are omitted | Physical F |
49
- | 3 | The Level 2 information boundary, with appearance reconstruction additionally evaluated | Physical F and appearance R |
 
 
 
 
50
 
51
- The levels use the same underlying task physics and interaction implementation. Select a level with `level=1`, `level=2` or `level=3` in `act2sim.open_task(...)` and `act2sim.evaluate(...)`, or with the `--level` argument.
 
 
 
 
52
 
53
- Cloth can be opened, explored and reconstructed as before. Evaluating it with this revision fails before the submission is sealed, with an error naming the task and the data revision. The report status is `runtime_error` and carries no score.
 
 
 
 
54
 
55
- ## Run locally
56
 
57
- The code repository is currently private. The code and installation-reference links in this card require repository access; public dataset downloads are separate and do not require a Hugging Face token.
58
 
59
- The complete scoring profile is Linux x86-64, Python 3.12 and the locked CUDA 12 dependency builds. Drawer uses CPU; peg uses CUDA.
60
 
61
- A compatible NVIDIA driver, LLVM/OptiX native rendering libraries and sufficient local resources remain host prerequisites. On Linux, complete the [native-library recipe](https://github.com/EricSanchezok/act2sim/blob/e3cdf88/docs/deployment.md#native-rendering-libraries) before running `doctor` or the example below.
 
 
62
 
63
  ```sh
64
  git clone https://github.com/EricSanchezok/act2sim.git
@@ -69,183 +73,63 @@ uv run act2sim doctor
69
  uv run python examples/local_agent.py --task drawer/open_stop --level 1 --device cpu --output results/drawer-l1
70
  ```
71
 
72
- Package `0.7.0` resolves this revision by default.
73
-
74
- The [executable example](https://github.com/EricSanchezok/act2sim/blob/e3cdf88/examples/local_agent.py) demonstrates interaction, reset, a source submission and local evaluation. Its fixed generic reconstruction is a workflow example, and its scores do not establish reconstruction quality. Replace that reconstruction with your own agent's source submission.
75
 
76
- Use `assembly/peg_insert` with `device="cuda:0"`. To select a different physical GPU, set `CUDA_VISIBLE_DEVICES` before starting Python and keep logical `cuda:0`.
77
-
78
- Reproducing a `v0.6.0` or `v0.5.0` score needs the code commit named in that revision's card. Package `0.7.0` rejects their calibrations with a calibration-identity error.
79
-
80
- ## Automatic downloads and offline use
81
-
82
- The package downloads the requested public files automatically. It resolves the selected Hub revision to an immutable commit and verifies manifest sizes and SHA-256 values. A token is not required. Downloads happen in stages:
83
- 1. Interaction fetches the target and robot assets.
84
- 2. Levels 1–2 evaluation adds the task's calibration files.
85
- 3. Level 3 adds the shared perceptual-model payload, about 2.2 GB.
86
-
87
- `HF_HOME` selects the cache, and `offline=True` uses verified cached data only. `ACT2SIM_DATA_ROOT` selects an already materialized tree, and `ACT2SIM_DATA_REPO` / `ACT2SIM_DATA_REVISION` select a mirror or immutable revision.
88
 
89
- ## Dataset layout and evidence
90
 
91
  ```text
92
- manifest.json
93
- README.md
94
- tasks/drawer/target/
95
- tasks/peg/target/
96
- tasks/shirt/target/
97
- robots/fr3/
98
- evaluation/drawer/
99
- evaluation/peg/
100
- calibration-inputs/
101
- calibration-inputs-lineage.json
102
- quality/cloth/
103
- metrics/
104
- native-runtime/linux-x86_64/
105
  ```
106
 
107
- [manifest.json](manifest.json) declares:
108
- - drawer and peg with `evaluation_root` and `metrics_root`;
109
- - cloth without them, which makes it interaction-only.
110
-
111
- | Task | Frozen protocol | Sealed calibration | Ceiling control |
112
- | --- | --- | --- | --- |
113
- | Drawer | [Protocol](evaluation/drawer/protocol.json) | [Calibration](evaluation/drawer/calibration.json) | [Ceiling](evaluation/drawer/ceiling.json) |
114
- | Peg | [Protocol](evaluation/peg/protocol.json) | [Calibration](evaluation/peg/calibration.json) | [Ceiling](evaluation/peg/ceiling.json) |
115
-
116
- `calibration-inputs/` keeps the `v0.5.0` reference and prior artifacts, which this revision reuses. Its lineage file still binds them.
117
- - `calibration-inputs/peg-panel-v0.7.0/` keeps the peg panel's records:
118
- - the 270 candidates with their pilot cohort;
119
- - the prior-only selection records and the selected panel;
120
- - the record of the revision run that sealed it, and its comparison with the validity study's calibration.
121
- - `calibration-inputs/drawer-panel-v0.7.0/` keeps the same records for the drawer panel.
122
-
123
- `quality/cloth/` is the `v0.5.0` cloth quality record. It describes the cloth target that is still published for interaction.
124
-
125
- ## Frozen evaluation configuration
126
-
127
- Both evaluated tasks use:
128
- - the 16 published prior members of `v0.5.0`;
129
- - the `v0.5.0` target reference;
130
- - eight held-out appearance views, plus the original public views (drawer three, peg three);
131
- - correspondence `part_owner_v1`.
132
-
133
- | Task | Probes (eligible) | ID / OOD / critical | Events (eligible) | Event term |
134
- | --- | --- | --- | --- | --- |
135
- | `drawer/open_stop` | 72 (70) | 24 / 24 / 24 | 216 (0) | `continuous_only` |
136
- | `assembly/peg_insert` | 60 (60) | 20 / 20 / 20 | 60 (0) | `continuous_only` |
137
-
138
- **Correspondence.** Every point of a rigid target part follows the submitted body that holds a strict majority of the part's points within a size-relative radius. Drawer parts use about 32 mm and peg parts about 3.4 mm. A submission that splits a part across bodies is judged by the majority owner, and the split is reported.
139
-
140
- **Eligibility.** A probe enters F only if its prior spread is at least the spread produced by one Level 1 pose-noise observation (ρ ≥ 1). The decision uses prior replays and pose-noise draws only. Excluded probes stay in reports as sanity checks.
141
-
142
- **Both tasks score trajectories only.** Every event thresholds a quantity of its own probe's trajectory, such as how far the drawer opened or whether the peg is still held. The continuous term already scores that trajectory. In the validity study, a single flipped event moved a drawer score by 0.18–0.34. Both calibrations therefore declare their task continuous-only:
143
- - every event is recorded as ineligible with the reason `continuous_only`;
144
- - F = 1 − S/S_prior on the continuous term;
145
- - events and their outcomes remain in reports as diagnostics.
146
-
147
- **Drawer panel.** The 30 `v0.6.0` probes were kept, and 14 new probes per category were added by a prior-only rule. Candidates and the rule were frozen before any replay. Selection used a separate pilot prior cohort. The target was not used.
148
-
149
- **Peg physics (`peg-v2`).**
150
- - The FR3 fingers' sliding friction is 0.01 in the precision profile. MuJoCo combines two shapes' friction by the larger value, so finger–peg contact now uses the peg's own coefficient. In `peg-v1`, the fingers' 1.0 hid it.
151
- - Only one finger is driven; the other follows through an equality constraint. That constraint is now stiffened to two solver substeps, so the grip does not yield.
152
- - The probes close the fingers fully from the grasp waypoint on.
153
- - The geometry, priors and robot effort limits are unchanged.
154
-
155
- **Peg panel.** 270 loosening candidates were generated and replayed on a separate pilot prior cohort, never on the target. 174 passed the prior-only selection rule. The panel keeps the 20 candidates per category with the largest prior-to-noise spread ratio among them. Each probe grasps the peg with the public qualification tape and lifts it. It then opens the fingers to a fixed aperture and holds or shakes. Whether and how fast the peg slips depends on its mass and its friction.
156
-
157
- **Ceiling control.** Eight pose-noise draws of the target reference, scored as submissions, give the F an agent with exact physics and one noisy Level 1 observation can reach. It is reported and is never used to normalize.
158
-
159
- | Task | Ceiling F mean (min–max) | Template-baseline F, Level 1 | Template-baseline R, Level 3 |
160
- | --- | --- | --- | --- |
161
- | Drawer | 0.530 (0.029–0.779) | −4.804 | −0.606 |
162
- | Peg | 0.673 (0.397–0.837) | −7.032 | 0.169 |
163
-
164
- The drawer ceiling is lower than the `v0.6.0` value of 0.775 because the event term no longer lifts it. On the same 72 probes with the `v0.6.0` event rule, the ceiling was 0.747: the noisy observations matched the events almost perfectly (event skill 0.964), while their continuous skill was 0.530, as here. By category, the drawer ceiling means are 0.600 (id), 0.386 (ood) and 0.603 (critical). One ood draw scores below zero: the eligibility rule admits probes whose single-observation spread is as large as the prior spread, so one unlucky observation can score as badly as a typical prior member on such a probe.
165
-
166
- By category, the peg ceiling means are 0.676 (id), 0.513 (ood) and 0.766 (critical).
167
 
168
- The `v0.6.0` card reported a peg ceiling of 0.808 for a different panel. That value was measured with a multiworld replay defect, corrected in this code revision: every batched pose-noise draw started from the first draw's peg pose.
169
 
170
- F aggregates losses before prior normalization. R is one minus mean LPIPS divided by the calibrated gray-control mean LPIPS. Scores may be negative. Read the report's validity status alongside its numbers.
171
 
172
- ## Validity evidence
 
 
 
173
 
174
- Drawer and peg scoring in this revision were checked by preregistered studies: [spec 0013](https://github.com/EricSanchezok/act2sim/blob/e3cdf88/docs/specs/0013-rigid-f-validity-round-2.md) for drawer and [spec 0014](https://github.com/EricSanchezok/act2sim/blob/e3cdf88/docs/specs/0014-peg-v2-validity.md) for peg.
175
- - Instances share the drawer geometry and vary its hidden physical parameters.
176
- - Candidate reconstructions carry known parameter errors.
177
- - An independent behavior measure, built from actions outside the panel, ranks the candidates.
178
- - The study then asks whether F ranks them the same way, stably and without rewarding memorized trajectories.
179
 
180
- On six newly sealed instances, scored once after the rules were committed:
 
 
181
 
182
- | Measure | `v0.7.0` drawer | `v0.6.0` drawer, same instances | Required |
183
- | --- | --- | --- | --- |
184
- | Pairs ordered as the behavior measure orders them | 0.907 (95% CI 0.841–0.940) | 0.894 | ≥ 0.90 |
185
- | Lowest per-class ordering | 0.879 (moving mass) | — | ≥ 0.80 |
186
- | Ranking agreement between half-panels, 5th percentile | 0.948 | 0.869 (minimum 0.779) | ≥ 0.90 |
187
 
188
- Prior resampling and the memorization check also passed.
189
 
190
- Peg was checked the same way twice, on four newly sealed instances each time:
191
- 1. The first confirmation used a 102-probe panel and passed every testable criterion (ordering 0.962, half-panel 5th percentile 0.931).
192
- 2. The released 60-probe panel was then chosen by a fixed rule and confirmed on four further instances. Its outcome was fixed in advance: had it failed, the 102-probe panel would have been released instead.
193
 
194
- | Measure | `v0.7.0` peg, 60 probes | `peg-v1`, round 1 sealed instances | Required |
195
  | --- | --- | --- | --- |
196
- | Pairs ordered as the behavior measure orders them | 0.972 (95% CI 0.965–0.981) | 0.767 | ≥ 0.90 |
197
- | Lowest per-class ordering (friction, mass, structure, memorizer) | 1.00 | — | ≥ 0.80 |
198
- | Representation variants ordered correctly | 1.00 | — | ≥ 0.90 |
199
- | Ranking agreement between half-panels, 5th percentile | 0.958 | not stable | ≥ 0.90 |
200
- | Memorization check | 1.00 | — | ≥ 0.90 |
201
-
202
- The release calibration reproduces the peg calibration of that confirmation exactly: reference, prior and ceiling trajectories and every probe's eligibility are identical.
203
 
204
- ## Source and numerical identity
205
 
206
- The numerical implementation for this revision is [Act2Sim commit e3cdf88](https://github.com/EricSanchezok/act2sim/commit/e3cdf88). Calibration code SHA-256: `38e6463b5a61903b071a650f5457a523ca6953ddfe01f264a2dc143cbef89142`. Peg prior, pose-noise and ceiling replays ran as multiworld batches over four GPUs (`scripts/validation/batched.py`), which reproduces the single-world replay bit for bit.
207
-
208
- **Calibration host:**
209
- - Ubuntu 22.04 x86-64 with RTX 4090 GPUs and driver 570.153.02.
210
- - Drawer was sealed on CPU with `llvm_ad_rgb`.
211
- - Peg was sealed on CUDA with `cuda_ad_rgb`, using OptiX from the matching NVIDIA driver package.
212
-
213
- Numerical compatibility uses:
214
- - the calibration digest;
215
- - the backend family;
216
- - the required renderer, for Level 3;
217
- - the pinned numerical dependency identities.
218
-
219
- ## Licenses and provenance
220
 
221
- Unchanged from `v0.5.0`. This collection retains multiple upstream licenses, and Act2Sim's MIT code license does not relicense the task assets, robot assets or pretrained models.
222
 
223
- | Asset | Upstream source and revision | License identified by the asset notice | Dataset notice |
224
- | --- | --- | --- | --- |
225
- | Drawer | [RoboCasa](https://github.com/robocasa/robocasa), `3443cdf3dd20f9ae8ed41adba442059b506891d0`; OpenDrawer episode `20250819/episode_000000` | CC-BY-4.0 | [Drawer NOTICE](tasks/drawer/target/NOTICE.md) |
226
- | Peg/hole | [IsaacLab Factory](https://github.com/isaac-sim/IsaacLab), `fd4b00bbf07a667bdadac8a4d5bd8c64b5a57a4a` | BSD-3-Clause | [Peg NOTICE](tasks/peg/target/NOTICE.md) |
227
- | Shirt | [DexGarmentLab](https://huggingface.co/datasets/wayrise/DexGarmentLab), `2ba4092676006bc98c257e7b822de39526fd9692` | Apache-2.0, as identified in the upstream dataset card | [Shirt NOTICE](tasks/shirt/target/NOTICE.md) |
228
- | Act2Sim FR3 robot pack | [Newton assets, Franka/Panda source](https://github.com/newton-physics/newton-assets/tree/f8fb7abcbeba2318814a74f3eeb02780ad7925d6/franka_emika_panda), `f8fb7abcbeba2318814a74f3eeb02780ad7925d6` | Apache-2.0 | [Robot NOTICE](robots/fr3/NOTICE.md) |
229
-
230
- [metrics/NOTICE.json](metrics/NOTICE.json) and [metrics/weights.json](metrics/weights.json) retain the perceptual-model sources, revisions, hashes and license notes.
231
-
232
- The optional [native LLVM runtime attachment](native-runtime/linux-x86_64/receipt.json) and its [copyright notices](native-runtime/linux-x86_64/COPYRIGHT) are unchanged. NVIDIA driver and OptiX libraries are not redistributed.
233
-
234
- ## Scientific limitations and intended use
235
-
236
- - **Peg measures mass and friction through slip.**
237
- - Every probe loosens a grasp, so peg F sees mass and friction as they affect slip.
238
- - It sees contact geometry poorly. Tilt probes were dropped because they penalized equivalent representations of the peg.
239
- - When the peg's friction is high, the fully closed grip already holds it, so overestimating a high friction costs little F.
240
- - Pushes of the standing peg were measured as noise-dominated and are not in the panel.
241
- - **Peg outcomes contain chaotic landings.** A peg that slips lands upright or lying. Selection kept only candidates whose cross-GPU difference was at most half the pose-noise spread. Untested GPU models may still differ more.
242
- - **The ceiling is below 1 by construction.** It measures one noisy observation with exact physics, and it shows how much of the scale is reachable. The drawer ceiling varies widely between single observations (0.03–0.78).
243
- - **Drawer critical events count only through trajectories.** A reconstruction that misses a critical event, such as the drawer reaching its stop, loses F only through the trajectory distance this causes.
244
- - **Statistical coverage is limited.** There are 16 prior members and four bases with eight draws each. The validity studies used six drawer and twice four peg confirmation instances, all of one geometry per task. Their candidates were synthetic, with known errors, not real agent submissions.
245
- - **Cloth is not evaluable in this revision.** Its `v0.5.0` quality findings still describe the published target.
246
-
247
- Report the following when sharing results:
248
- - task, level, data and protocol identities;
249
- - runtime identity;
250
- - reconstruction method;
251
- - validity status.
 
1
  ---
2
+ pretty_name: Act2Sim v0.7.0 Public Task Data
3
+ viewer: false
4
  license: other
5
  license_name: mixed-upstream-licenses
6
  license_link: https://huggingface.co/datasets/SII-EricSanchez/act2sim-data/blob/v0.7.0/README.md#licenses-and-provenance
 
13
  - robotics
14
  ---
15
 
16
+ # Act2Sim Public Task Data (`v0.7.0`)
17
 
18
+ Versioned benchmark assets, sealed evaluation protocols, and perceptual metric weights for **[Act2Sim](https://github.com/EricSanchezok/act2sim)** (`0.7.0`).
 
 
 
 
 
 
 
19
 
20
+ > **Note:** This repository is a **manifest-indexed asset store** consumed automatically by the `act2sim` Python package—not a `datasets.load_dataset(...)` image-classification dataset.
21
 
22
+ ---
 
 
 
23
 
24
+ ## Code vs. Data Boundary
25
 
26
+ | Repository | Responsibility |
27
+ | --- | --- |
28
+ | **[`EricSanchezok/act2sim`](https://github.com/EricSanchezok/act2sim)** | Python SDK (`open_task`, `Session`, `evaluate`), Newton/MuJoCo simulation runtime, CLI, and benchmark documentation |
29
+ | **`SII-EricSanchez/act2sim-data`** (this repo) | `manifest.json` (SHA-256 verified), target scenes (`tasks/`), FR3 robot pack (`robots/`), sealed calibration protocols (`evaluation/`), perceptual weights (`metrics/`), and provenance records |
30
 
31
+ ---
32
 
33
+ ## What Changed in `v0.7.0`
 
 
 
 
34
 
35
+ - **`drawer/open_stop` (`drawer-v1`)**: 72-probe panel, trajectory-only physical scoring (`continuous_only`, no event term).
36
+ - **`assembly/peg_insert` (`peg-v2`)**: Robot fingers now read the peg's own friction with a stiffened mimic coupling (`equality_timeconst_substeps=2`); 60-probe loosening panel; trajectory-only physical scoring (`continuous_only`) sensitive to both mass and friction.
37
+ - **`cloth/unfold_shirt` (`shirt-v2`)**: Unchanged from `v0.6.0` (interaction-only; no calibrated evaluation in `v0.7.0`).
38
+
39
+ > Drawer and peg scores from `v0.6.0` and `v0.7.0` are **not comparable**. Compare results only within matching task and protocol identities.
40
+
41
+ ---
42
+
43
+ ## Tasks & Information Levels
44
 
45
+ | Task | Revision | World | Robots | Evaluation (`v0.7.0`) | Scoring Device |
46
+ | --- | --- | --- | --- | --- | --- |
47
+ | `drawer/open_stop` | `drawer-v1` | Passive articulated drawer | 1× FR3 | Levels 1–3 | CPU |
48
+ | `assembly/peg_insert` | `peg-v2` | 8 mm peg-and-hole assembly | 1× FR3 | Levels 1–3 | Linux CUDA (`cuda:0`) |
49
+ | `cloth/unfold_shirt` | `shirt-v2` | Deformable shirt | 2× FR3 | Interaction only | — |
50
 
51
+ | Level | Observations Provided to Agent | Reported Metrics |
52
+ | --- | --- | --- |
53
+ | **1** | RGB-D, robot proprioception, exterior geometry, noisy whole-object poses | Physical `F` |
54
+ | **2** | RGB-D and robot proprioception (no object geometry or poses) | Physical `F` |
55
+ | **3** | Same as Level 2, with appearance reconstruction evaluated | Physical `F` + Appearance `R` |
56
 
57
+ ---
58
 
59
+ ## Quickstart (Automatic Download)
60
 
61
+ You do **not** need to clone this repository manually or pass a Hugging Face token. `act2sim` downloads and SHA-256-verifies only the files needed for each stage:
62
 
63
+ 1. **Interaction**: target & FR3 robot assets
64
+ 2. **Levels 1–2 Evaluation**: task calibration protocol & reference trajectories
65
+ 3. **Level 3 Evaluation**: shared perceptual-model weights (~2.2 GB)
66
 
67
  ```sh
68
  git clone https://github.com/EricSanchezok/act2sim.git
 
73
  uv run python examples/local_agent.py --task drawer/open_stop --level 1 --device cpu --output results/drawer-l1
74
  ```
75
 
76
+ Environment overrides: `HF_HOME` (cache path), `ACT2SIM_DATA_ROOT` (local unpacked snapshot), `ACT2SIM_DATA_REPO` / `ACT2SIM_DATA_REVISION`, and `offline=True` (verified cache only). On Linux, complete the [native-library recipe](https://github.com/EricSanchezok/act2sim/blob/e3cdf88/docs/deployment.md#native-rendering-libraries) first; for `assembly/peg_insert`, keep logical `cuda:0` and select physical GPUs via `CUDA_VISIBLE_DEVICES`.
 
 
77
 
78
+ ---
 
 
 
 
 
 
 
 
 
 
 
79
 
80
+ ## Repository Layout
81
 
82
  ```text
83
+ manifest.json # Authoritative file inventory, sizes, and SHA-256 digests
84
+ tasks/{drawer,peg,shirt}/target/ # Public target scene assets and upstream NOTICE files
85
+ robots/fr3/ # Shared FR3 robot pack
86
+ evaluation/{drawer,peg}/ # Frozen protocols, priors, ceiling controls, and views
87
+ metrics/ # Shared perceptual metric weights (LPIPS / CLIP / DINO / LoRA)
88
+ calibration-inputs/ # Panel selection & calibration provenance records
89
+ calibration-inputs-lineage.json # Lineage binding v0.5.0 priors/references to v0.7.0
90
+ quality/cloth/ # v0.5.0 cloth target quality diagnostics
91
+ native-runtime/linux-x86_64/ # Optional LLVM runtime attachment and notices
 
 
 
 
92
  ```
93
 
94
+ ---
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
95
 
96
+ ## Calibration & Validity Summary
97
 
98
+ Both evaluated rigid tasks use `part_owner_v1` correspondence, 16 prior members from `v0.5.0`, 8 held-out views + 3 public views, and `continuous_only` F scoring (`F = 1 - S / S_prior`; negative scores preserved). `R` is `1 - mean_LPIPS / gray_control_mean_LPIPS`.
99
 
100
+ | Task | Probes (Eligible) | ID / OOD / Critical | Ceiling `F` Mean (Min–Max) | Template Baseline `F` (L1) | Template Baseline `R` (L3) |
101
+ | --- | --- | --- | --- | --- | --- |
102
+ | `drawer/open_stop` | 72 (70) | 24 / 24 / 24 | 0.530 (0.029–0.779) | −4.804 | −0.606 |
103
+ | `assembly/peg_insert` | 60 (60) | 20 / 20 / 20 | 0.673 (0.397–0.837) | −7.032 | 0.169 |
104
 
105
+ Preregistered validity studies ([spec 0013](https://github.com/EricSanchezok/act2sim/blob/e3cdf88/docs/specs/0013-rigid-f-validity-round-2.md) for drawer, [spec 0014](https://github.com/EricSanchezok/act2sim/blob/e3cdf88/docs/specs/0014-peg-v2-validity.md) for peg):
106
+ - **Drawer (`v0.7.0`)**: Pairwise ordering `0.907` (95% CI `0.841–0.940`), lowest per-class `0.879` (moving mass), half-panel 5th percentile `0.948`.
107
+ - **Peg (`v0.7.0`, 60 probes)**: Pairwise ordering `0.972` (95% CI `0.965–0.981`), lowest per-class `1.00`, representation variants `1.00`, half-panel 5th percentile `0.958`, memorization check `1.00`.
 
 
108
 
109
+ **Source & Numerical Identity:**
110
+ - Code commit: [`e3cdf88`](https://github.com/EricSanchezok/act2sim/commit/e3cdf88) · Calibration code SHA-256: `38e6463b5a61903b071a650f5457a523ca6953ddfe01f264a2dc143cbef89142`
111
+ - Calibration host: Ubuntu 22.04 x86-64, RTX 4090 (driver `570.153.02`), `llvm_ad_rgb` (drawer CPU) / `cuda_ad_rgb` + OptiX (peg CUDA).
112
 
113
+ ---
 
 
 
 
114
 
115
+ ## Licenses & Provenance
116
 
117
+ This collection retains multiple upstream licenses; Act2Sim's MIT code license does not relicense task assets, robot assets, or pretrained models.
 
 
118
 
119
+ | Asset | Upstream Source & Revision | License | Dataset Notice |
120
  | --- | --- | --- | --- |
121
+ | Drawer | [RoboCasa](https://github.com/robocasa/robocasa) `3443cdf3dd20f9ae8ed41adba442059b506891d0` (`OpenDrawer` `20250819/episode_000000`) | CC-BY-4.0 | [Drawer NOTICE](tasks/drawer/target/NOTICE.md) |
122
+ | Peg / hole | [IsaacLab Factory](https://github.com/isaac-sim/IsaacLab) `fd4b00bbf07a667bdadac8a4d5bd8c64b5a57a4a` | BSD-3-Clause | [Peg NOTICE](tasks/peg/target/NOTICE.md) |
123
+ | Shirt | [DexGarmentLab](https://huggingface.co/datasets/wayrise/DexGarmentLab) `2ba4092676006bc98c257e7b822de39526fd9692` | Apache-2.0 | [Shirt NOTICE](tasks/shirt/target/NOTICE.md) |
124
+ | FR3 robot pack | [Newton assets](https://github.com/newton-physics/newton-assets/tree/f8fb7abcbeba2318814a74f3eeb02780ad7925d6/franka_emika_panda) `f8fb7abcbeba2318814a74f3eeb02780ad7925d6` | Apache-2.0 | [Robot NOTICE](robots/fr3/NOTICE.md) |
 
 
 
125
 
126
+ See [metrics/NOTICE.json](metrics/NOTICE.json), [metrics/weights.json](metrics/weights.json), [native-runtime/linux-x86_64/receipt.json](native-runtime/linux-x86_64/receipt.json), and [native-runtime/linux-x86_64/COPYRIGHT](native-runtime/linux-x86_64/COPYRIGHT) for metric and runtime notices.
127
 
128
+ ---
 
 
 
 
 
 
 
 
 
 
 
 
 
129
 
130
+ ## Scientific Limitations
131
 
132
+ - **Peg measures mass and friction through slip**; contact geometry is weakly probed, high friction values above the closed-grip threshold cost little `F`, and chaotic landing outcomes (upright vs. lying) are filtered for RTX 4090 / 5090 consistency only.
133
+ - **Ceiling is below 1 by construction** (single noisy Level 1 observation with exact physics; drawer ceiling spans `0.03–0.78`).
134
+ - **Drawer events** affect `F` only through trajectory distance; **cloth** is interaction-only in `v0.7.0`.
135
+ - **Statistical coverage** is limited to 16 prior members, 4 bases × 8 draws, and synthetic confirmation candidates (6 drawer, 2 × 4 peg).