Datasets:
RoboRACE: code, cached agent outputs, results, media, SO-101 renders
Browse files- .gitattributes +0 -6
- README.md +219 -61
.gitattributes
CHANGED
|
@@ -6,9 +6,3 @@
|
|
| 6 |
*.webp binary
|
| 7 |
*.mp4 binary
|
| 8 |
*.pdf binary
|
| 9 |
-
media/figures/racenet_scaling.png filter=lfs diff=lfs merge=lfs -text
|
| 10 |
-
models/racenet/candidates/v1_kaggle/model.safetensors filter=lfs diff=lfs merge=lfs -text
|
| 11 |
-
models/racenet/candidates/v1_local/model.safetensors filter=lfs diff=lfs merge=lfs -text
|
| 12 |
-
models/racenet/candidates/v2_local/model.safetensors filter=lfs diff=lfs merge=lfs -text
|
| 13 |
-
viewer/erqa_hosted_agents/test.parquet filter=lfs diff=lfs merge=lfs -text
|
| 14 |
-
models/racenet_m/model.safetensors filter=lfs diff=lfs merge=lfs -text
|
|
|
|
| 6 |
*.webp binary
|
| 7 |
*.mp4 binary
|
| 8 |
*.pdf binary
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
README.md
CHANGED
|
@@ -40,6 +40,22 @@ configs:
|
|
| 40 |
path: viewer/panda_views/evaluation.parquet
|
| 41 |
- config_name: panda_vlm_scores
|
| 42 |
data_files: viewer/panda_vlm_scores/train.parquet
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 43 |
- config_name: trackD_maze_results
|
| 44 |
data_files: viewer/trackD_maze_results/train.parquet
|
| 45 |
- config_name: maze_llm_agents
|
|
@@ -113,6 +129,8 @@ The GIF is a 12-second highlight; the full 43-second film (3 episodes) is the <a
|
|
| 113 |
| `data/so101/episodes.jsonl` | per-episode metadata: task, family (put / stack), scripted mode, **physics ground truth**, final object pose |
|
| 114 |
| `results/so101/vlm_scores.jsonl` | Cosmos-Reason2-2B P(success) for every (episode, camera) — config **`so101_vlm_scores`** |
|
| 115 |
| `data/panda/...`, `results/panda/vlm_scores.jsonl` | the same for the **Franka Panda** replication (220 episodes × 5 cameras) — configs **`panda_views`**, **`panda_vlm_scores`** |
|
|
|
|
|
|
|
| 116 |
| `results/trackD/rows.json` | Track D maze swarm: SPL and success of every fusion rule in every (maze, attack, liar-count) cell — config **`trackD_maze_results`** |
|
| 117 |
| `results/maze_llm/<agent>.jsonl` | LLM agents on text mazes: every attempt, validator feedback, repair count — config **`maze_llm_agents`** |
|
| 118 |
| `models/racenet/` | RACE-Net weights (safetensors), config and training log; also at [Nabidnur/RoboRACE-Net](https://huggingface.co/Nabidnur/RoboRACE-Net) |
|
|
@@ -149,11 +167,12 @@ pip install -r requirements.txt && python scripts/run_benchmark.py --reuse-track
|
|
| 149 |
| **ERQA** (Gemini Robotics team, 400 Q) | reproduce Cosmos-Reason2-2B on CPU | NVIDIA reports 37.75 | **33.8** (95% CI 29.3–38.5) ✓ inside |
|
| 150 |
| **ERQA** multi-agent | 4 VLM agents, 283 shared Q | best single agent 54.1 | best swarm **54.8** · robot's own model 37.1 → **43.8** with CGA |
|
| 151 |
| **PlanBench** Blocksworld | plan generation, 92 instances | official one-shot prompt | LLM-Modulo + swarm cascade: see [table](#planbench-blocksworld--the-validator-is-the-planner) |
|
| 152 |
-
| **Track A** SO-101 camera swarm | success verification under 7 attacks | receiver alone 71.3 · majority 59.8 | **
|
| 153 |
| **Track A-Panda** cross-embodiment | same protocol on a Franka Panda, methods frozen | receiver alone 74.5 · majority 62.2 | **87.0** race_s_online (new) |
|
| 154 |
-
| **RACE-Net** (our model, 0.2 M params) | learned fusion, zero-shot on real camera swarms | RACE-S online 82.7 / 87.0 | **
|
|
|
|
| 155 |
| **Track B** 12-robot rendezvous | mission success, up to 5 Byzantine robots | W-MSR 82.1 | **94.3** RACE-C + W-MSR (rank 1) |
|
| 156 |
-
| **Track D** maze swarm (
|
| 157 |
| **Track C** LLM planners | exact-plan accuracy | zero-shot mean 88.1 | engineered mean **98.2** · flash-lite 69.0 → **92.9** |
|
| 158 |
|
| 159 |
<p align="center"><img src="https://huggingface.co/datasets/Nabidnur/RoboRACE-Bench/resolve/main/media/figures/overview.png" width="920"></p>
|
|
@@ -162,7 +181,7 @@ pip install -r requirements.txt && python scripts/run_benchmark.py --reuse-track
|
|
| 162 |
[SO-101 camera swarm](https://huggingface.co/datasets/Nabidnur/RoboRACE-Bench/resolve/main/media/videos/robot_rollouts.mp4) ·
|
| 163 |
[Byzantine rendezvous](https://huggingface.co/datasets/Nabidnur/RoboRACE-Bench/resolve/main/media/videos/swarm_rendezvous.mp4) ·
|
| 164 |
[PlanBench LLM-Modulo replay](https://huggingface.co/datasets/Nabidnur/RoboRACE-Bench/resolve/main/media/videos/planbench_modulo.mp4) ·
|
| 165 |
-
[maze swarm](https://huggingface.co/datasets/Nabidnur/RoboRACE-Bench/resolve/main/media/videos/maze_swarm.mp4) ·
|
| 166 |
[Franka Panda camera swarm](https://huggingface.co/datasets/Nabidnur/RoboRACE-Bench/resolve/main/media/videos/panda_rollouts.mp4)
|
| 167 |
|
| 168 |
## The idea in one paragraph
|
|
@@ -186,6 +205,24 @@ clean-data gains), **RACE-C + W-MSR** (continuous robot consensus) and an **engi
|
|
| 186 |
Exact numbers come from the checkpoint: 2.44 B parameters (LLM decoder 1,409 M, embeddings 311 M, untied head 311 M, ViT 302 M, merger + DeepStack 101 M).
|
| 187 |
Vector version: <a href="https://huggingface.co/datasets/Nabidnur/RoboRACE-Bench/resolve/main/media/architecture/plotnn/cosmos_verifier_plotnn.pdf">PDF</a> (LaTeX source alongside). Diagrams are drawn by <code>scripts/make_architecture.py</code> in the style of Sebastian Raschka's LLM architecture figures.</sub></p>
|
| 188 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 189 |
---
|
| 190 |
|
| 191 |
## Part I — public benchmarks
|
|
@@ -271,14 +308,14 @@ near-miss placements and wrong objects. Each of 5 cameras is an agent (Cosmos-Re
|
|
| 271 |
|
| 272 |
| rank (90% CI) | method | mean | clean | liars < 50% | liars ≥ 50% | worst cell |
|
| 273 |
|---|---|---:|---:|---:|---:|---:|
|
| 274 |
-
| **1** (1–1) | **
|
| 275 |
-
| **2** (2–2) | **
|
| 276 |
-
| **3** (3–3) | **
|
| 277 |
-
| **4** (4–
|
| 278 |
-
| **5** (
|
| 279 |
-
| **6** (5–7) |
|
| 280 |
-
| **7** (6–
|
| 281 |
-
| **8** (
|
| 282 |
| ref | REF stack_lr (labelled) | **82.6** | 92.7 | 84.9 | 78.1 | 48.2 |
|
| 283 |
|
| 284 |
<p align="center"><img src="https://huggingface.co/datasets/Nabidnur/RoboRACE-Bench/resolve/main/media/figures/trackA_vs_liars.png" width="980"></p>
|
|
@@ -300,13 +337,13 @@ physics-based labels were then generated with a **Franka Panda** (7-DOF, MuJoCo
|
|
| 300 |
| rank (90% CI) | method | mean | clean | liars < 50% | liars ≥ 50% | worst cell |
|
| 301 |
|---|---|---:|---:|---:|---:|---:|
|
| 302 |
| **1** (1–1) | **race_s_online (new)** | **87.0** | 91.8 | 88.7 | 83.6 | 53.6 |
|
| 303 |
-
| **2** (2–2) | **
|
| 304 |
-
| **3** (3–
|
| 305 |
-
| **
|
| 306 |
-
| **5** (
|
| 307 |
-
| **
|
| 308 |
-
| **7** (
|
| 309 |
-
| **8** (
|
| 310 |
| ref | REF stack_lr (labelled) | **84.9** | 96.4 | 87.0 | 80.7 | 50.0 |
|
| 311 |
|
| 312 |
<table><tr><td width="50%"><img src="https://huggingface.co/datasets/Nabidnur/RoboRACE-Bench/resolve/main/media/figures/cross_embodiment.png"></td><td width="50%"><img src="https://huggingface.co/datasets/Nabidnur/RoboRACE-Bench/resolve/main/media/figures/panda_gallery.png"></td></tr></table>
|
|
@@ -318,56 +355,119 @@ majority vote 48.8). The left panel plots every method's mean accuracy on the SO
|
|
| 318 |
<p align="center"><img src="https://huggingface.co/datasets/Nabidnur/RoboRACE-Bench/resolve/main/media/figures/trackAP_leaderboard.png" width="760"></p>
|
| 319 |
</details>
|
| 320 |
|
| 321 |
-
### RACE-Net —
|
| 322 |
|
| 323 |
-
RACE and RACE-S run a hand-specified EM on every window. **RACE-Net** is a
|
| 324 |
*learned* the same receiver-anchored inference: it reads a whole window of unlabelled swarm reports, knows only which column is the receiver,
|
| 325 |
-
and outputs P(success)
|
| 326 |
-
and then applied unchanged to the real Cosmos-Reason2 camera swarms.
|
|
|
|
| 327 |
|
| 328 |
<table><tr><td width="42%"><img src="https://huggingface.co/datasets/Nabidnur/RoboRACE-Bench/resolve/main/media/architecture/racenet.png"></td><td width="58%"><img src="https://huggingface.co/datasets/Nabidnur/RoboRACE-Bench/resolve/main/media/figures/racenet.png"></td></tr></table>
|
| 329 |
|
| 330 |
-
| mean accuracy over all attack cells | A0 procedural (development) | SO-101 (test) | Franka Panda (
|
| 331 |
|---|---:|---:|---:|
|
| 332 |
| majority vote | 68.5 | 59.8 | 62.2 |
|
| 333 |
| Dawid–Skene (crowd-kit) | 69.4 | 60.1 | 65.0 |
|
| 334 |
| RACE online | 90.3 | 77.7 | 83.2 |
|
| 335 |
| RACE-S online | 91.2 | 82.7 | 87.0 |
|
| 336 |
-
| **RACE-Net (learned, zero-shot)** | **
|
|
|
|
| 337 |
| *labelled stacker (reference)* | 84.8 | 82.6 | 84.9 |
|
| 338 |
|
| 339 |
-
Liar-majority cells: RACE-Net
|
| 340 |
(RACE-S online: 79.8 / 83.6).
|
| 341 |
Weights, config and a standalone inference file are in [`models/racenet`](models/racenet) and on the Hub:
|
| 342 |
[🤗 Nabidnur/RoboRACE-Net](https://huggingface.co/Nabidnur/RoboRACE-Net).
|
| 343 |
|
| 344 |
-
**Checkpoints and selection.** Rule
|
| 345 |
-
|
| 346 |
-
|
| 347 |
-
|
| 348 |
-
|---|---|---:|---:|---:|---:|---:|---:|
|
| 349 |
-
| **v1 · local CPU · 6k steps (released)** | v1 | 6,000 | 93.4 / 96.8 | **90.4** | 81.6 | 81.8 | 81.8 |
|
| 350 |
-
| v1 · Kaggle CPU · 15k steps | v1 | 15,000 | 93.9 / 97.2 | **90.2** | 80.9 | 82.3 | 84.5 |
|
| 351 |
-
| v2 · local CPU · 7k steps | v2 | 7,000 | 92.6 / 96.9 | **90.2** | 82.6 | 83.5 | 88.5 |
|
| 352 |
|
| 353 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 354 |
|
| 355 |
-
|
| 356 |
|
| 357 |
<p align="center"><img src="https://huggingface.co/datasets/Nabidnur/RoboRACE-Bench/resolve/main/media/figures/racenet_scaling.png" width="980"></p>
|
| 358 |
|
| 359 |
**Scaling study** (`scripts/racenet_analysis.py`; an independent synthetic generator, not the training distribution; 20 seeds per point):
|
| 360 |
|
| 361 |
-
- **Liar majority.** With 3 of 5 agents lying: RACE-Net
|
| 362 |
-
majority 45.5. With 15 of 16 lying:
|
| 363 |
-
- **Short windows.** With only 8 unlabelled episodes: RACE-Net
|
| 364 |
no label-free method beats the receiver before about 32 episodes.
|
| 365 |
-
- **Large swarms.** At 24 agents (larger than any training swarm): RACE-Net
|
| 366 |
RACE-S online 93.2.
|
| 367 |
-
- **Weak receiver (open weakness).** If the receiver is only 55% accurate and nobody lies, RACE-Net
|
| 368 |
-
|
| 369 |
-
|
| 370 |
-
- **
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 371 |
|
| 372 |
### Track B — rendezvous with Byzantine robots
|
| 373 |
|
|
@@ -412,27 +512,76 @@ step on its fused belief map. Up to four robots corrupt their broadcasts (false
|
|
| 412 |
that walls off the goal, sleepers). Score = SPL (success weighted by path length) of the honest robots.
|
| 413 |
|
| 414 |
<p align="center"><img src="https://huggingface.co/datasets/Nabidnur/RoboRACE-Bench/resolve/main/media/videos/maze_swarm.gif" width="960"><br><sub>Three of six robots invert their reports. Majority map fusion believes the false walls (red);
|
| 415 |
-
RACE
|
| 416 |
two rules; averages are in the table. <a href="https://huggingface.co/datasets/Nabidnur/RoboRACE-Bench/resolve/main/media/videos/maze_swarm.mp4">MP4</a></sub></p>
|
| 417 |
|
|
|
|
|
|
|
|
|
|
|
|
|
| 418 |
| rank (90% CI) | method | mean | clean | liars < 50% | liars ≥ 50% | worst cell |
|
| 419 |
|---|---|---:|---:|---:|---:|---:|
|
| 420 |
-
| **1** (1–1) | **
|
| 421 |
-
| **2** (2–2) |
|
| 422 |
-
| **3** (3–3) |
|
| 423 |
-
| **4** (4–
|
| 424 |
-
| **5** (
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 425 |
| ref | REF oracle | **71.3** | 76.8 | 74.5 | 67.5 | 0.0 |
|
| 426 |
|
| 427 |
-
|
| 428 |
-
|
| 429 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
| 430 |
|
| 431 |
<p align="center"><img src="https://huggingface.co/datasets/Nabidnur/RoboRACE-Bench/resolve/main/media/figures/trackD_vs_liars.png" width="980"></p>
|
| 432 |
|
| 433 |
-
|
| 434 |
-
|
| 435 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 436 |
|
| 437 |
<details><summary><b>More Track D figures</b></summary>
|
| 438 |
<p align="center"><img src="https://huggingface.co/datasets/Nabidnur/RoboRACE-Bench/resolve/main/media/figures/trackD_leaderboard.png" width="640"></p>
|
|
@@ -478,13 +627,19 @@ the models are newer, hosted agents were run on quota-limited subsets with a bri
|
|
| 478 |
| Dawid–Skene implemented correctly | ≥90% decision agreement with the crowd-kit reference implementation on identical data |
|
| 479 |
| RACE-S calibration is rank-invariant | a cubic distortion of one agent changes ≤2% of decisions |
|
| 480 |
| SO-101 labels are physical | labels read from the simulator after settling; replay of a cached seed must reproduce its label |
|
| 481 |
-
| no test-set tuning | RACE-S hyper-parameters fixed on synthetic data; CGA / sign-only reported as ablations on *every* benchmark; Track D tuned on separate development mazes; RACE-Net trained on
|
| 482 |
| findings replicate on a new robot | all methods frozen before the Franka Panda data was generated (Track A-Panda) |
|
| 483 |
-
| RACE-Net
|
| 484 |
| maze checker is exact | all 40 BFS-shortest paths validate with their optimal length (`tests/test_maze.py`) |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 485 |
|
| 486 |
|
| 487 |
-
`pytest` runs
|
| 488 |
|
| 489 |
## Reproduce (CPU)
|
| 490 |
|
|
@@ -530,13 +685,15 @@ src/robotics/vlm/ cosmos_reason.py (logit P(yes) verifier)
|
|
| 530 |
scripts/ run_benchmark · train_racenet · select_racenet · racenet_analysis · make_figures · make_videos · make_gifs · make_architecture · make_plotnn · make_readme ·
|
| 531 |
publish_hf + make_space + make_playground (Hugging Face dataset, model, Spaces) · agents
|
| 532 |
results/ LEADERBOARD.md · leaderboard.json · cached agent outputs
|
|
|
|
| 533 |
media/ figures · videos (mp4 + gif) · architecture (Raschka-style + PlotNeuralNet)
|
| 534 |
```
|
| 535 |
|
| 536 |
## Limitations
|
| 537 |
- Track A renders are simulated. The SO-101 and Panda meshes are real, but lighting and textures are synthetic, and all cameras share one VLM.
|
| 538 |
- Track D is a grid world with exact local sensing; the 3-D view in the video is a rendering of the same run, not a physics simulation.
|
| 539 |
-
- RACE-Net is trained on a hand-written simulator of swarms
|
|
|
|
| 540 |
- Online variants fit on the current unlabelled evaluation window; a deployed robot would use a sliding window.
|
| 541 |
- ERQA swarm numbers cover the questions that every hosted agent answered within free-tier quotas. Groq Qwen and Gemma-4 were throttled to near-zero coverage and are excluded.
|
| 542 |
- PlanBench uses a seeded 100-instance subset. Track C saturates after engineering.
|
|
@@ -545,7 +702,8 @@ media/ figures · videos (mp4 + gif) · architecture (Raschka
|
|
| 545 |
## Provenance & citation
|
| 546 |
RACE / *Liars Are Information* ([public demo Space](https://huggingface.co/spaces/Nabidnur/liars-are-information-demo); the dataset `Nabidnur/liars-are-information` is private) ·
|
| 547 |
AIP-AD ([`Nabidnur/AIP-AD`](https://huggingface.co/Nabidnur/AIP-AD)) · ERQA (Gemini Robotics team) · PlanBench (Valmeekam et al., NeurIPS 2023 D&B) ·
|
| 548 |
-
LLM-Modulo (Kambhampati et al., ICML 2024) · W-MSR (LeBlanc et al., IEEE JSAC 2013) ·
|
|
|
|
| 549 |
Cosmos-Reason2 (NVIDIA) · SO-101 model (MuJoCo Menagerie) · PlotNeuralNet (H. Iqbal) · diagram style after S. Raschka.
|
| 550 |
|
| 551 |
```bibtex
|
|
|
|
| 40 |
path: viewer/panda_views/evaluation.parquet
|
| 41 |
- config_name: panda_vlm_scores
|
| 42 |
data_files: viewer/panda_vlm_scores/train.parquet
|
| 43 |
+
- config_name: so101_holdout_views
|
| 44 |
+
data_files:
|
| 45 |
+
- split: history
|
| 46 |
+
path: viewer/so101_holdout_views/history.parquet
|
| 47 |
+
- split: evaluation
|
| 48 |
+
path: viewer/so101_holdout_views/evaluation.parquet
|
| 49 |
+
- config_name: so101_holdout_vlm_scores
|
| 50 |
+
data_files: viewer/so101_holdout_vlm_scores/train.parquet
|
| 51 |
+
- config_name: panda_holdout_views
|
| 52 |
+
data_files:
|
| 53 |
+
- split: history
|
| 54 |
+
path: viewer/panda_holdout_views/history.parquet
|
| 55 |
+
- split: evaluation
|
| 56 |
+
path: viewer/panda_holdout_views/evaluation.parquet
|
| 57 |
+
- config_name: panda_holdout_vlm_scores
|
| 58 |
+
data_files: viewer/panda_holdout_vlm_scores/train.parquet
|
| 59 |
- config_name: trackD_maze_results
|
| 60 |
data_files: viewer/trackD_maze_results/train.parquet
|
| 61 |
- config_name: maze_llm_agents
|
|
|
|
| 129 |
| `data/so101/episodes.jsonl` | per-episode metadata: task, family (put / stack), scripted mode, **physics ground truth**, final object pose |
|
| 130 |
| `results/so101/vlm_scores.jsonl` | Cosmos-Reason2-2B P(success) for every (episode, camera) — config **`so101_vlm_scores`** |
|
| 131 |
| `data/panda/...`, `results/panda/vlm_scores.jsonl` | the same for the **Franka Panda** replication (220 episodes × 5 cameras) — configs **`panda_views`**, **`panda_vlm_scores`** |
|
| 132 |
+
| `data/*_holdout/...`, `results/*_holdout/vlm_scores.jsonl`, `results/holdout/FREEZE.json` | the **hold-out**: 170 new episodes per arm, evaluated only after the recorded freeze commit — configs **`so101_holdout_views`**, **`panda_holdout_views`** and the matching `*_vlm_scores` |
|
| 133 |
+
| `models/racenet`, `models/racenet_m` | weights of both models (**RACE-Net-V**, **RACE-Net-M**) and every candidate checkpoint |
|
| 134 |
| `results/trackD/rows.json` | Track D maze swarm: SPL and success of every fusion rule in every (maze, attack, liar-count) cell — config **`trackD_maze_results`** |
|
| 135 |
| `results/maze_llm/<agent>.jsonl` | LLM agents on text mazes: every attempt, validator feedback, repair count — config **`maze_llm_agents`** |
|
| 136 |
| `models/racenet/` | RACE-Net weights (safetensors), config and training log; also at [Nabidnur/RoboRACE-Net](https://huggingface.co/Nabidnur/RoboRACE-Net) |
|
|
|
|
| 167 |
| **ERQA** (Gemini Robotics team, 400 Q) | reproduce Cosmos-Reason2-2B on CPU | NVIDIA reports 37.75 | **33.8** (95% CI 29.3–38.5) ✓ inside |
|
| 168 |
| **ERQA** multi-agent | 4 VLM agents, 283 shared Q | best single agent 54.1 | best swarm **54.8** · robot's own model 37.1 → **43.8** with CGA |
|
| 169 |
| **PlanBench** Blocksworld | plan generation, 92 instances | official one-shot prompt | LLM-Modulo + swarm cascade: see [table](#planbench-blocksworld--the-validator-is-the-planner) |
|
| 170 |
+
| **Track A** SO-101 camera swarm | success verification under 7 attacks | receiver alone 71.3 · majority 59.8 | **83.4** race_net (new, learned) (rank 1) |
|
| 171 |
| **Track A-Panda** cross-embodiment | same protocol on a Franka Panda, methods frozen | receiver alone 74.5 · majority 62.2 | **87.0** race_s_online (new) |
|
| 172 |
+
| **RACE-Net-V** (our model, 0.2 M params) | learned fusion, zero-shot on real camera swarms | RACE-S online 82.7 / 87.0 | **83.4** SO-101 (rank 1) / **84.2** Panda · RACE-Pool **82.0** / **85.9** |
|
| 173 |
+
| **Hold-out** (new episodes, both arms, evaluated after the freeze) | same protocol, nothing tuned on it | receiver alone 72.7 / 77.1 · majority 62.3 / 64.2 | RACE-S online **82.6** / **85.4** · RACE-Pool **81.2** / **86.7** · RACE-Net-V 79.0 / 84.3 |
|
| 174 |
| **Track B** 12-robot rendezvous | mission success, up to 5 Byzantine robots | W-MSR 82.1 | **94.3** RACE-C + W-MSR (rank 1) |
|
| 175 |
+
| **Track D** maze swarm · **RACE-Net-M** (our model, 45 K params) | SPL under 6 attacks, 10 fusion rules | W-MSR trimming 64.1 · beta-reputation 63.8 · EigenTrust 63.0 | **70.2** RACE-Net-M (rank 1; oracle 71.3) · hand-written RACE map 66.2 |
|
| 176 |
| **Track C** LLM planners | exact-plan accuracy | zero-shot mean 88.1 | engineered mean **98.2** · flash-lite 69.0 → **92.9** |
|
| 177 |
|
| 178 |
<p align="center"><img src="https://huggingface.co/datasets/Nabidnur/RoboRACE-Bench/resolve/main/media/figures/overview.png" width="920"></p>
|
|
|
|
| 181 |
[SO-101 camera swarm](https://huggingface.co/datasets/Nabidnur/RoboRACE-Bench/resolve/main/media/videos/robot_rollouts.mp4) ·
|
| 182 |
[Byzantine rendezvous](https://huggingface.co/datasets/Nabidnur/RoboRACE-Bench/resolve/main/media/videos/swarm_rendezvous.mp4) ·
|
| 183 |
[PlanBench LLM-Modulo replay](https://huggingface.co/datasets/Nabidnur/RoboRACE-Bench/resolve/main/media/videos/planbench_modulo.mp4) ·
|
| 184 |
+
[maze swarm](https://huggingface.co/datasets/Nabidnur/RoboRACE-Bench/resolve/main/media/videos/maze_swarm.mp4) · [maze, wall-inventing liars](https://huggingface.co/datasets/Nabidnur/RoboRACE-Bench/resolve/main/media/videos/maze_false_walls.mp4) ·
|
| 185 |
[Franka Panda camera swarm](https://huggingface.co/datasets/Nabidnur/RoboRACE-Bench/resolve/main/media/videos/panda_rollouts.mp4)
|
| 186 |
|
| 187 |
## The idea in one paragraph
|
|
|
|
| 205 |
Exact numbers come from the checkpoint: 2.44 B parameters (LLM decoder 1,409 M, embeddings 311 M, untied head 311 M, ViT 302 M, merger + DeepStack 101 M).
|
| 206 |
Vector version: <a href="https://huggingface.co/datasets/Nabidnur/RoboRACE-Bench/resolve/main/media/architecture/plotnn/cosmos_verifier_plotnn.pdf">PDF</a> (LaTeX source alongside). Diagrams are drawn by <code>scripts/make_architecture.py</code> in the style of Sebastian Raschka's LLM architecture figures.</sub></p>
|
| 207 |
|
| 208 |
+
### Our two models: one design
|
| 209 |
+
|
| 210 |
+
<table><tr>
|
| 211 |
+
<td width="50%"><img src="https://huggingface.co/datasets/Nabidnur/RoboRACE-Bench/resolve/main/media/architecture/racenet.png"></td>
|
| 212 |
+
<td width="50%"><img src="https://huggingface.co/datasets/Nabidnur/RoboRACE-Bench/resolve/main/media/architecture/racenet_m.png"></td>
|
| 213 |
+
</tr></table>
|
| 214 |
+
|
| 215 |
+
| | **RACE-Net-V** · pick-and-place | **RACE-Net-M** · maze |
|
| 216 |
+
|---|---|---|
|
| 217 |
+
| question | did the arm succeed? (camera swarm) | which cells are walls? (robot swarm) |
|
| 218 |
+
| receiver's anchor | its own score for every episode | its own exact sensing of some cells |
|
| 219 |
+
| structure | axial set transformer over agents × episodes | anchored channel encoder → peer transformer → signed read-out, 3 refinement rounds |
|
| 220 |
+
| shared design | anchor on the receiver · estimate each peer's channel in context · **invert** consistent liars · sets, not sequences · synthetic training only |
|
| 221 |
+
| size · training | 213,889 parameters · synthetic swarms | 44,837 parameters · simulated mazes with random liars |
|
| 222 |
+
|
| 223 |
+
Both come from RACE and AIP in the *Liars Are Information* program: RACE supplies the anchored, signed channel estimation; AIP supplies the pooling of
|
| 224 |
+
experts (RACE-Pool). The full description is in [docs/MODELS.md](https://huggingface.co/datasets/Nabidnur/RoboRACE-Bench/blob/main/docs/MODELS.md).
|
| 225 |
+
|
| 226 |
---
|
| 227 |
|
| 228 |
## Part I — public benchmarks
|
|
|
|
| 308 |
|
| 309 |
| rank (90% CI) | method | mean | clean | liars < 50% | liars ≥ 50% | worst cell |
|
| 310 |
|---|---|---:|---:|---:|---:|---:|
|
| 311 |
+
| **1** (1–1) | **race_net (new, learned)** | **83.4** | 88.2 | 85.1 | 80.0 | 53.6 |
|
| 312 |
+
| **2** (2–2) | **race_s_online (new)** | **82.7** | 87.3 | 84.2 | 79.8 | 54.5 |
|
| 313 |
+
| **3** (3–3) | **race_pool (new)** | **82.0** | 86.2 | 83.8 | 78.7 | 54.5 |
|
| 314 |
+
| **4** (4–4) | **race_online (new)** | **77.7** | 83.6 | 79.7 | 73.9 | 49.1 |
|
| 315 |
+
| **5** (5–6) | **race CGA (new)** | **77.1** | 83.6 | 79.7 | 72.1 | 39.1 |
|
| 316 |
+
| **6** (5–7) | **race_s (new)** | **76.3** | 85.1 | 78.6 | 71.9 | 49.1 |
|
| 317 |
+
| **7** (6–8) | race_online sign-only anchor (ablation) | **76.0** | 83.6 | 79.6 | 69.1 | 16.4 |
|
| 318 |
+
| **8** (7–8) | race | **75.0** | 83.3 | 78.0 | 69.3 | 44.5 |
|
| 319 |
| ref | REF stack_lr (labelled) | **82.6** | 92.7 | 84.9 | 78.1 | 48.2 |
|
| 320 |
|
| 321 |
<p align="center"><img src="https://huggingface.co/datasets/Nabidnur/RoboRACE-Bench/resolve/main/media/figures/trackA_vs_liars.png" width="980"></p>
|
|
|
|
| 337 |
| rank (90% CI) | method | mean | clean | liars < 50% | liars ≥ 50% | worst cell |
|
| 338 |
|---|---|---:|---:|---:|---:|---:|
|
| 339 |
| **1** (1–1) | **race_s_online (new)** | **87.0** | 91.8 | 88.7 | 83.6 | 53.6 |
|
| 340 |
+
| **2** (2–2) | **race_pool (new)** | **85.9** | 90.5 | 87.7 | 82.4 | 53.6 |
|
| 341 |
+
| **3** (3–3) | **race_net (new, learned)** | **84.2** | 86.0 | 85.6 | 81.5 | 54.5 |
|
| 342 |
+
| **4** (4–4) | **race_online (new)** | **83.2** | 90.0 | 85.9 | 78.0 | 49.1 |
|
| 343 |
+
| **5** (5–6) | **race CGA (new)** | **82.7** | 90.0 | 85.9 | 76.6 | 34.5 |
|
| 344 |
+
| **6** (6–8) | race_online sign-only anchor (ablation) | **81.8** | 90.0 | 85.9 | 74.2 | 5.5 |
|
| 345 |
+
| **7** (5–7) | race | **81.7** | 94.2 | 86.8 | 71.9 | 5.5 |
|
| 346 |
+
| **8** (7–8) | **race_s (new)** | **80.4** | 91.3 | 82.9 | 75.3 | 50.0 |
|
| 347 |
| ref | REF stack_lr (labelled) | **84.9** | 96.4 | 87.0 | 80.7 | 50.0 |
|
| 348 |
|
| 349 |
<table><tr><td width="50%"><img src="https://huggingface.co/datasets/Nabidnur/RoboRACE-Bench/resolve/main/media/figures/cross_embodiment.png"></td><td width="50%"><img src="https://huggingface.co/datasets/Nabidnur/RoboRACE-Bench/resolve/main/media/figures/panda_gallery.png"></td></tr></table>
|
|
|
|
| 355 |
<p align="center"><img src="https://huggingface.co/datasets/Nabidnur/RoboRACE-Bench/resolve/main/media/figures/trackAP_leaderboard.png" width="760"></p>
|
| 356 |
</details>
|
| 357 |
|
| 358 |
+
### RACE-Net-V — the learned verification model
|
| 359 |
|
| 360 |
+
RACE and RACE-S run a hand-specified EM on every window. **RACE-Net-V** is a 214,145-parameter axial set transformer that has
|
| 361 |
*learned* the same receiver-anchored inference: it reads a whole window of unlabelled swarm reports, knows only which column is the receiver,
|
| 362 |
+
and outputs P(success) for every episode. It is trained **only on synthetic swarms** (no benchmark data, labels or agent outputs)
|
| 363 |
+
and then applied unchanged to the real Cosmos-Reason2 camera swarms. The released checkpoint is the EM-refinement variant: it also receives the RACE-S and RACE estimates of the window and learns to refine them.
|
| 364 |
+
**RACE-Pool** is the AIP-style pool of RACE-Net-V, RACE-S and RACE ([design](https://huggingface.co/datasets/Nabidnur/RoboRACE-Bench/blob/main/docs/MODELS.md)).
|
| 365 |
|
| 366 |
<table><tr><td width="42%"><img src="https://huggingface.co/datasets/Nabidnur/RoboRACE-Bench/resolve/main/media/architecture/racenet.png"></td><td width="58%"><img src="https://huggingface.co/datasets/Nabidnur/RoboRACE-Bench/resolve/main/media/figures/racenet.png"></td></tr></table>
|
| 367 |
|
| 368 |
+
| mean accuracy over all attack cells | A0 procedural (development) | SO-101 (test) | Franka Panda (test) |
|
| 369 |
|---|---:|---:|---:|
|
| 370 |
| majority vote | 68.5 | 59.8 | 62.2 |
|
| 371 |
| Dawid–Skene (crowd-kit) | 69.4 | 60.1 | 65.0 |
|
| 372 |
| RACE online | 90.3 | 77.7 | 83.2 |
|
| 373 |
| RACE-S online | 91.2 | 82.7 | 87.0 |
|
| 374 |
+
| **RACE-Net-V (learned, zero-shot)** | **91.5** | **83.4** | **84.2** |
|
| 375 |
+
| **RACE-Pool (RACE-Net-V + RACE-S + RACE)** | **92.4** | **82.0** | **85.9** |
|
| 376 |
| *labelled stacker (reference)* | 84.8 | 82.6 | 84.9 |
|
| 377 |
|
| 378 |
+
Liar-majority cells: RACE-Net-V 80.0 on SO-101 and 81.5 on the Panda
|
| 379 |
(RACE-S online: 79.8 / 83.6).
|
| 380 |
Weights, config and a standalone inference file are in [`models/racenet`](models/racenet) and on the Hub:
|
| 381 |
[🤗 Nabidnur/RoboRACE-Net](https://huggingface.co/Nabidnur/RoboRACE-Net).
|
| 382 |
|
| 383 |
+
**Checkpoints and selection.** Rule: highest mean accuracy of RACE-Net-V over three development sets (A0 evaluation window, SO-101 history split, Panda history split; equal weight); ties within 0.1 point go to the higher synthetic accuracy. The history splits are the benchmark's unlabelled calibration
|
| 384 |
+
data; no label-free method ever used their labels, so they serve as development data for the two arms, and the evaluation
|
| 385 |
+
splits stay test sets. Test columns were computed after the choice and are shown for transparency. On the same development
|
| 386 |
+
sets RACE-S online scores 84.9 and RACE online 83.8.
|
|
|
|
|
|
|
|
|
|
|
|
|
| 387 |
|
| 388 |
+
| checkpoint | generator · type | steps | params | synthetic (v3) | dev: RACE-Net-V (selects) | dev: RACE-Pool | test SO-101 / Panda: RACE-Net-V | test: RACE-Pool |
|
| 389 |
+
|---|---|---:|---:|---:|---:|---:|---:|---:|
|
| 390 |
+
| **v3 refine · local · 6k (released)** | v3 · EM-refinement | 6,000 | 214,145 | 97.8 | **84.0** | 85.6 | 83.4 / 84.2 | 82.0 / 85.9 |
|
| 391 |
+
| v3 large · Kaggle · 9k | v3 · plain | 9,000 | 589,729 | 97.9 | 83.8 | 85.6 | 83.6 / 84.1 | 82.2 / 85.9 |
|
| 392 |
+
| v3 · Kaggle · 15k · seed 1 | v3 · plain | 15,000 | 213,889 | 97.9 | 83.7 | 85.5 | 83.6 / 83.6 | 82.0 / 85.9 |
|
| 393 |
+
| v2 · local · 7k | v2 · plain | 7,000 | 213,889 | 96.7 | 83.4 | 85.6 | 82.6 / 83.5 | 81.9 / 86.1 |
|
| 394 |
+
| v3 · Kaggle · 15k · seed 0 | v3 · plain | 15,000 | 213,889 | 97.9 | 83.3 | 85.6 | 83.8 / 84.0 | 81.9 / 86.1 |
|
| 395 |
+
| v1 · local · 6k | v1 · plain | 6,000 | 213,889 | 97.6 | 82.2 | 85.0 | 81.6 / 81.8 | 81.7 / 85.5 |
|
| 396 |
+
| v1 · Kaggle · 15k | v1 · plain | 15,000 | 213,889 | 97.8 | 82.0 | 85.0 | 80.9 / 82.3 | 81.4 / 85.4 |
|
| 397 |
|
| 398 |
+
Checkpoints marked Kaggle were trained in Kaggle CPU notebooks (4 vCPU; logs in `results/kaggle/`). All checkpoints are in `models/racenet/candidates/`.
|
| 399 |
|
| 400 |
<p align="center"><img src="https://huggingface.co/datasets/Nabidnur/RoboRACE-Bench/resolve/main/media/figures/racenet_scaling.png" width="980"></p>
|
| 401 |
|
| 402 |
**Scaling study** (`scripts/racenet_analysis.py`; an independent synthetic generator, not the training distribution; 20 seeds per point):
|
| 403 |
|
| 404 |
+
- **Liar majority.** With 3 of 5 agents lying: RACE-Net-V 80.1, RACE-S online 77.6, RACE online 76.8,
|
| 405 |
+
majority 45.5. With 15 of 16 lying: 86.7 / 83.4 / 78.7 / 21.9.
|
| 406 |
+
- **Short windows.** With only 8 unlabelled episodes: RACE-Net-V 71.7, RACE-S online 72.8, receiver alone 70.0;
|
| 407 |
no label-free method beats the receiver before about 32 episodes.
|
| 408 |
+
- **Large swarms.** At 24 agents (larger than any training swarm): RACE-Net-V 93.8, RACE online 93.4,
|
| 409 |
RACE-S online 93.2.
|
| 410 |
+
- **Weak receiver (open weakness).** If the receiver is only 55% accurate and nobody lies, RACE-Net-V reaches 65.8 and RACE-S online
|
| 411 |
+
54.4, while majority reaches 81.8 and discrete RACE 82.4. Score-level anchoring trusts a weak
|
| 412 |
+
receiver too much; from a 70% receiver upwards the gap is gone.
|
| 413 |
+
- **Against the first release.** Weak receivers: 65.8 at 55% and 87.3 at 70% receiver accuracy (first release 57.1 and 77.4). 24-agent swarms: 93.8 (first release 89.1).
|
| 414 |
+
- **RACE-Pool** on the same sweeps: 79.2 with 3 of 5 lying, 94.9 at 24 agents, 68.2 with a 55% receiver.
|
| 415 |
+
|
| 416 |
+
### Hold-out — a test nothing was tuned on
|
| 417 |
+
|
| 418 |
+
Both earlier test tracks were looked at while RACE-Net-V was being developed. For a clean final number, **1,700 new camera
|
| 419 |
+
views** were generated with new seeds for both arms (60 history + 110 evaluation episodes each). The views were scored by the frozen Cosmos-Reason2
|
| 420 |
+
pipeline, and no fusion method was run on them until commit [`5b2bc8e`](https://github.com/Hisernberg/robotics/commit/5b2bc8e89da83f8893f2bf99e962b9312ded18b4) had frozen every model,
|
| 421 |
+
rule and threshold (`results/holdout/FREEZE.json`; `run_benchmark.py` refuses to evaluate the hold-out without that record).
|
| 422 |
+
|
| 423 |
+
<table><tr><td width="50%">
|
| 424 |
+
|
| 425 |
+
**SO-101 hold-out**
|
| 426 |
+
|
| 427 |
+
| rank (90% CI) | method | mean | clean | liars < 50% | liars ≥ 50% | worst cell |
|
| 428 |
+
|---|---|---:|---:|---:|---:|---:|
|
| 429 |
+
| **1** (1–1) | **race_s_online (new)** | **82.6** | 86.7 | 84.1 | 79.7 | 53.6 |
|
| 430 |
+
| **2** (2–2) | **race_pool (new)** | **81.2** | 84.9 | 82.9 | 78.0 | 52.7 |
|
| 431 |
+
| **3** (3–3) | **race_online (new)** | **79.9** | 86.5 | 82.1 | 75.6 | 52.7 |
|
| 432 |
+
| **4** (4–5) | **race CGA (new)** | **79.4** | 86.5 | 82.1 | 74.3 | 37.3 |
|
| 433 |
+
| **5** (4–6) | **race_net (new, learned)** | **79.0** | 77.1 | 80.0 | 77.2 | 55.5 |
|
| 434 |
+
| **6** (5–6) | race_online sign-only anchor (ablation) | **78.4** | 86.5 | 82.2 | 71.3 | 11.8 |
|
| 435 |
+
| **7** (7–7) | race | **74.3** | 80.0 | 76.8 | 69.7 | 33.6 |
|
| 436 |
+
| ref | REF stack_lr (labelled) | **82.6** | 93.6 | 85.3 | 77.5 | 45.5 |
|
| 437 |
+
|
| 438 |
+
</td><td width="50%">
|
| 439 |
+
|
| 440 |
+
**Franka Panda hold-out**
|
| 441 |
+
|
| 442 |
+
| rank (90% CI) | method | mean | clean | liars < 50% | liars ≥ 50% | worst cell |
|
| 443 |
+
|---|---|---:|---:|---:|---:|---:|
|
| 444 |
+
| **1** (1–1) | **race_pool (new)** | **86.7** | 90.9 | 88.7 | 83.1 | 56.4 |
|
| 445 |
+
| **2** (2–3) | **race_online (new)** | **85.9** | 92.7 | 89.0 | 80.1 | 51.8 |
|
| 446 |
+
| **3** (2–4) | **race_s_online (new)** | **85.4** | 89.1 | 87.0 | 82.5 | 58.2 |
|
| 447 |
+
| **4** (3–5) | **race CGA (new)** | **84.8** | 92.7 | 88.9 | 77.1 | 35.5 |
|
| 448 |
+
| **5** (4–6) | **race_net (new, learned)** | **84.3** | 87.5 | 85.7 | 81.6 | 54.5 |
|
| 449 |
+
| **6** (5–7) | race_online sign-only anchor (ablation) | **83.4** | 92.7 | 88.9 | 73.0 | 6.4 |
|
| 450 |
+
| **7** (6–7) | race | **82.1** | 89.5 | 85.3 | 76.1 | 48.2 |
|
| 451 |
+
| ref | REF stack_lr (labelled) | **82.6** | 93.6 | 84.7 | 78.5 | 50.0 |
|
| 452 |
+
|
| 453 |
+
</td></tr></table>
|
| 454 |
+
|
| 455 |
+
| hold-out | comparison | difference | 95% CI (cells) | better / worse |
|
| 456 |
+
|---|---|---:|---|---:|
|
| 457 |
+
| SO-101 | race_pool − race_s_online | -1.4 | -1.8 … -1.0 | 110 / 199 |
|
| 458 |
+
| SO-101 | race_net − race_s_online | -3.6 | -4.2 … -3.0 | 65 / 254 |
|
| 459 |
+
| SO-101 | race_net − race_online | -0.9 | -1.6 … -0.2 | 148 / 191 |
|
| 460 |
+
| Panda | race_pool − race_s_online | +1.3 | +1.0 … +1.5 | 270 / 49 |
|
| 461 |
+
| Panda | race_net − race_s_online | -1.2 | -1.5 … -0.8 | 80 / 225 |
|
| 462 |
+
| Panda | race_net − race_online | -1.6 | -2.3 … -0.9 | 81 / 252 |
|
| 463 |
+
|
| 464 |
+
**Reading.** On data nothing was tuned on, the receiver-anchored family keeps its margin over every literature baseline
|
| 465 |
+
(majority 62.3 / 64.2, Dawid–Skene (crowd-kit) 62.3 / 68.8, receiver alone
|
| 466 |
+
72.7 / 77.1). Among our own methods the EM-based ones are ahead: RACE-S online is rank 1 on SO-101
|
| 467 |
+
and RACE-Pool is rank 1 on the Panda. The learned **RACE-Net-V is rank 5 and 5**
|
| 468 |
+
(79.0 / 84.3); its first place on the SO-101 test track (83.4) does **not** replicate here. For camera swarms
|
| 469 |
+
the recommendation is therefore RACE-S online or RACE-Pool. RACE-Net-V is within 1 to 4 points of them, which is the honest size of the gap between
|
| 470 |
+
learning the inference from simulated swarms and fitting it on the window at hand.
|
| 471 |
|
| 472 |
### Track B — rendezvous with Byzantine robots
|
| 473 |
|
|
|
|
| 512 |
that walls off the goal, sleepers). Score = SPL (success weighted by path length) of the honest robots.
|
| 513 |
|
| 514 |
<p align="center"><img src="https://huggingface.co/datasets/Nabidnur/RoboRACE-Bench/resolve/main/media/videos/maze_swarm.gif" width="960"><br><sub>Three of six robots invert their reports. Majority map fusion believes the false walls (red);
|
| 515 |
+
RACE-Net-M verifies each peer against cells the robot sensed itself and <b>inverts</b> the liars. This is the maze with the largest gap between the
|
| 516 |
two rules; averages are in the table. <a href="https://huggingface.co/datasets/Nabidnur/RoboRACE-Bench/resolve/main/media/videos/maze_swarm.mp4">MP4</a></sub></p>
|
| 517 |
|
| 518 |
+
**RACE-Net-M** is the learned map-fusion model ([design](https://huggingface.co/datasets/Nabidnur/RoboRACE-Bench/blob/main/docs/MODELS.md)): 44,837 parameters, trained on simulated maze swarms with *random* two-coin liars.
|
| 519 |
+
The benchmark's attack names never appear in its training, and the colluding ring attack is outside that family. It is compared with six rules from the
|
| 520 |
+
literature and with our hand-written `race_map` on 20 fresh mazes × 25 conditions.
|
| 521 |
+
|
| 522 |
| rank (90% CI) | method | mean | clean | liars < 50% | liars ≥ 50% | worst cell |
|
| 523 |
|---|---|---:|---:|---:|---:|---:|
|
| 524 |
+
| **1** (1–1) | **race_net_m + reach-repair** | **70.2** | 76.8 | 73.1 | 66.8 | 0.0 |
|
| 525 |
+
| **2** (2–2) | **race_net_m** | **69.8** | 76.8 | 72.6 | 66.4 | 0.0 |
|
| 526 |
+
| **3** (3–3) | **race_map** | **66.2** | 76.8 | 69.7 | 61.8 | 0.0 |
|
| 527 |
+
| **4** (4–5) | wmsr_trim | **64.1** | 76.8 | 65.6 | 61.5 | 13.9 |
|
| 528 |
+
| **5** (4–6) | reputation | **63.8** | 76.8 | 67.9 | 58.5 | 0.0 |
|
| 529 |
+
| **6** (5–7) | eigentrust | **63.0** | 75.0 | 66.8 | 58.2 | 0.0 |
|
| 530 |
+
| **7** (6–8) | dawid_skene | **62.0** | 76.3 | 67.4 | 55.5 | 0.0 |
|
| 531 |
+
| **8** (7–8) | solo | **61.6** | 61.5 | 62.0 | 61.2 | 13.9 |
|
| 532 |
+
| **9** (9–9) | trust_all | **56.7** | 76.8 | 59.9 | 51.7 | 0.0 |
|
| 533 |
+
| **10** (10–10) | majority | **55.7** | 76.8 | 61.5 | 48.1 | 0.0 |
|
| 534 |
| ref | REF oracle | **71.3** | 76.8 | 74.5 | 67.5 | 0.0 |
|
| 535 |
|
| 536 |
+
- **Rank 1.** RACE-Net-M scores 69.8 (70.2 with reachability repair), against 66.2 for the hand-written `race_map` and 64.1 for the best
|
| 537 |
+
literature baseline (wmsr_trim). The oracle that knows who lies scores 71.3.
|
| 538 |
+
- **Goal reached in time:** RACE-Net-M + repair 97.8%, RACE-Net-M 97.4%, `race_map` 94.2%,
|
| 539 |
+
W-MSR trimming 96.5%, beta-reputation 91.9%, majority 82.1% (oracle 98.2%).
|
| 540 |
+
- **Wall-inventing liars**, the weak spot of `race_map` (59.5), are handled: RACE-Net-M 71.7.
|
| 541 |
+
- **Still open:** the colluding ring (66.2 with repair, oracle 71.0). The liars are honest everywhere
|
| 542 |
+
else, so their channel looks reliable until a robot stands next to the goal.
|
| 543 |
|
| 544 |
<p align="center"><img src="https://huggingface.co/datasets/Nabidnur/RoboRACE-Bench/resolve/main/media/figures/trackD_vs_liars.png" width="980"></p>
|
| 545 |
|
| 546 |
+
<p align="center"><img src="https://huggingface.co/datasets/Nabidnur/RoboRACE-Bench/resolve/main/media/videos/maze_false_walls.gif" width="960"><br><sub>Three of six robots invent walls. Beta-reputation believes dozens of them;
|
| 547 |
+
RACE-Net-M learns that these peers' wall reports are worthless while their "free" reports are true, and keeps only the latter.
|
| 548 |
+
<a href="https://huggingface.co/datasets/Nabidnur/RoboRACE-Bench/resolve/main/media/videos/maze_false_walls.mp4">MP4</a></sub></p>
|
| 549 |
+
|
| 550 |
+
Belief quality on held-out mazes (accuracy on cells a peer reported and the receiver has not sensed; 6,111 snapshots):
|
| 551 |
+
|
| 552 |
+
| model | all | 3 of 5 peers lie | 5 of 5 peers lie |
|
| 553 |
+
|---|---:|---:|---:|
|
| 554 |
+
| `race_map` (hand-written Bayesian rule) | 90.1 | 82.7 | 72.9 |
|
| 555 |
+
| rounds = 1 (no refinement) | 91.8 | 85.7 | 76.1 |
|
| 556 |
+
| rounds = 2 | 92.2 | 86.4 | 77.4 |
|
| 557 |
+
| rounds = 3 (released) | 92.3 | 86.7 | 77.6 |
|
| 558 |
+
| rounds = 3, second seed | 92.3 | 86.7 | 77.6 |
|
| 559 |
+
|
| 560 |
+
**Other swarm and map sizes.** The model was trained on 6 robots in 12×12-cell mazes. Unchanged, with half of the robots lying
|
| 561 |
+
(10 new mazes × 5 attacks per row; computed in a Kaggle CPU notebook):
|
| 562 |
+
|
| 563 |
+
<p align="center"><img src="https://huggingface.co/datasets/Nabidnur/RoboRACE-Bench/resolve/main/media/figures/racenet_m_scaling.png" width="960"></p>
|
| 564 |
+
|
| 565 |
+
| setting | liars | beta-reputation | W-MSR trimming | race_map | RACE-Net-M + repair | oracle |
|
| 566 |
+
|---|---:|---:|---:|---:|---:|---:|
|
| 567 |
+
| 4 robots · 8×8 | 2 | 74.4 | 72.0 | 73.3 | **73.6** | 76.1 |
|
| 568 |
+
| 6 robots · 12×12 (training size) | 3 | 61.1 | 54.0 | 63.6 | **67.7** | 66.9 |
|
| 569 |
+
| 10 robots · 12×12 | 5 | 70.7 | 60.5 | 70.6 | **76.7** | 77.5 |
|
| 570 |
+
| 8 robots · 16×16 | 4 | 57.5 | 59.4 | 62.5 | **69.5** | 67.7 |
|
| 571 |
+
|
| 572 |
+
In small mazes (8×8) every rule is within a few points of the others: there is little to gain from peers. The gap grows with the map and the swarm.
|
| 573 |
+
|
| 574 |
+
**Reachability repair** (planner ablation, same 500 cells). When the fused map says the goal cannot be reached, the standard planner drops every peer;
|
| 575 |
+
the repair un-believes only the claimed walls that enclose the goal. It helps every rule against the colluding ring, but it hurts rules whose map is
|
| 576 |
+
already polluted with false walls, so it only pays off on top of an accurate fusion model:
|
| 577 |
+
|
| 578 |
+
| fusion rule | standard fallback | + reach-repair | collude: standard → repair | false_walls: standard → repair |
|
| 579 |
+
|---|---:|---:|---:|---:|
|
| 580 |
+
| majority | 55.7 | 49.2 | 63.2 → 66.2 | 57.6 → 45.5 |
|
| 581 |
+
| eigentrust | 63.0 | 61.0 | 62.6 → 65.0 | 53.8 → 42.2 |
|
| 582 |
+
| reputation | 63.8 | 61.8 | 62.7 → 66.2 | 57.0 → 44.8 |
|
| 583 |
+
| race_map | 66.2 | 65.6 | 63.2 → 66.2 | 59.5 → 54.3 |
|
| 584 |
+
| race_net_m | 69.8 | 70.2 | 63.2 → 66.2 | 71.7 → 71.2 |
|
| 585 |
|
| 586 |
<details><summary><b>More Track D figures</b></summary>
|
| 587 |
<p align="center"><img src="https://huggingface.co/datasets/Nabidnur/RoboRACE-Bench/resolve/main/media/figures/trackD_leaderboard.png" width="640"></p>
|
|
|
|
| 627 |
| Dawid–Skene implemented correctly | ≥90% decision agreement with the crowd-kit reference implementation on identical data |
|
| 628 |
| RACE-S calibration is rank-invariant | a cubic distortion of one agent changes ≤2% of decisions |
|
| 629 |
| SO-101 labels are physical | labels read from the simulator after settling; replay of a cached seed must reproduce its label |
|
| 630 |
+
| no test-set tuning | RACE-S hyper-parameters fixed on synthetic data; CGA / sign-only reported as ablations on *every* benchmark; Track D tuned on separate development mazes; both RACE-Net models trained on simulated swarms only and selected on development data |
|
| 631 |
| findings replicate on a new robot | all methods frozen before the Franka Panda data was generated (Track A-Panda) |
|
| 632 |
+
| RACE-Net-V and RACE-Net-M are set functions | outputs unchanged under any permutation of peers or items (`tests/test_racenet.py`, `tests/test_racenet_m.py`) |
|
| 633 |
| maze checker is exact | all 40 BFS-shortest paths validate with their optimal length (`tests/test_maze.py`) |
|
| 634 |
+
| hold-out was not peeked at | `results/holdout/FREEZE.json` records the commit that froze every model and rule; `run_benchmark.py` skips the hold-out without it |
|
| 635 |
+
| RACE-Net-M uses liars instead of discarding them | a full inverter alone recovers the map, a coin-flipper adds no evidence (`tests/test_racenet_m.py`) |
|
| 636 |
+
| literature baselines are faithful | EigenTrust drops the inverter, trimming removes F votes per cell, Dawid–Skene is right up to a global flip (`tests/test_racenet_m.py`) |
|
| 637 |
+
| differences are not noise | paired bootstrap intervals (over mazes for Track D) and exact sign tests in the [leaderboard](https://huggingface.co/datasets/Nabidnur/RoboRACE-Bench/blob/main/results/LEADERBOARD.md) |
|
| 638 |
+
| maze data hygiene | five disjoint sets of maze seeds: training, validation, development, benchmark, scaling study ([protocol](https://huggingface.co/datasets/Nabidnur/RoboRACE-Bench/blob/main/docs/BENCHMARK.md)) |
|
| 639 |
+
| independent replication on Kaggle CPU | a Kaggle notebook (4 vCPU, Intel(R) Xeon(R) CPU @ 2.20GHz) re-ran the tests (53 passed, 4 skipped in 66.52s (0:01:06)) and recomputed the leaderboard from the public Hugging Face copy (Nabidnur/RoboRACE-Bench as of 2026-11-20 16:24 UTC): 1,217 leaderboard values compared, max abs difference 0.0000, 109.2 min |
|
| 640 |
|
| 641 |
|
| 642 |
+
`pytest` runs 64 tests. See [docs/BENCHMARK.md](https://huggingface.co/datasets/Nabidnur/RoboRACE-Bench/blob/main/docs/BENCHMARK.md) for the full protocol and threats to validity.
|
| 643 |
|
| 644 |
## Reproduce (CPU)
|
| 645 |
|
|
|
|
| 685 |
scripts/ run_benchmark · train_racenet · select_racenet · racenet_analysis · make_figures · make_videos · make_gifs · make_architecture · make_plotnn · make_readme ·
|
| 686 |
publish_hf + make_space + make_playground (Hugging Face dataset, model, Spaces) · agents
|
| 687 |
results/ LEADERBOARD.md · leaderboard.json · cached agent outputs
|
| 688 |
+
kaggle/ the two Kaggle CPU notebooks (RACE-Net training, independent replication)
|
| 689 |
media/ figures · videos (mp4 + gif) · architecture (Raschka-style + PlotNeuralNet)
|
| 690 |
```
|
| 691 |
|
| 692 |
## Limitations
|
| 693 |
- Track A renders are simulated. The SO-101 and Panda meshes are real, but lighting and textures are synthetic, and all cameras share one VLM.
|
| 694 |
- Track D is a grid world with exact local sensing; the 3-D view in the video is a rendering of the same run, not a physics simulation.
|
| 695 |
+
- RACE-Net-V is trained on a hand-written simulator of swarms. On the hold-out it is 1 to 4 points behind our EM methods (RACE-S online, RACE-Pool); its first place on the SO-101 test track did not replicate. A weak receiver (below about 60% accuracy) is still an open weakness of every score-level method here.
|
| 696 |
+
- RACE-Net-M is evaluated in one simulator (grid mazes, exact local sensing). A bloc that lies only around the goal still costs about 5 SPL points against the oracle.
|
| 697 |
- Online variants fit on the current unlabelled evaluation window; a deployed robot would use a sliding window.
|
| 698 |
- ERQA swarm numbers cover the questions that every hosted agent answered within free-tier quotas. Groq Qwen and Gemma-4 were throttled to near-zero coverage and are excluded.
|
| 699 |
- PlanBench uses a seeded 100-instance subset. Track C saturates after engineering.
|
|
|
|
| 702 |
## Provenance & citation
|
| 703 |
RACE / *Liars Are Information* ([public demo Space](https://huggingface.co/spaces/Nabidnur/liars-are-information-demo); the dataset `Nabidnur/liars-are-information` is private) ·
|
| 704 |
AIP-AD ([`Nabidnur/AIP-AD`](https://huggingface.co/Nabidnur/AIP-AD)) · ERQA (Gemini Robotics team) · PlanBench (Valmeekam et al., NeurIPS 2023 D&B) ·
|
| 705 |
+
LLM-Modulo (Kambhampati et al., ICML 2024) · W-MSR (LeBlanc et al., IEEE JSAC 2013) · EigenTrust (Kamvar et al., WWW 2003) ·
|
| 706 |
+
beta-reputation (Jøsang & Ismail, 2002) · SPL (Anderson et al., 2018) · Dawid & Skene (1979) · crowd-kit (Toloka) ·
|
| 707 |
Cosmos-Reason2 (NVIDIA) · SO-101 model (MuJoCo Menagerie) · PlotNeuralNet (H. Iqbal) · diagram style after S. Raschka.
|
| 708 |
|
| 709 |
```bibtex
|