PhysicalAI-US-Evaluation

A held-out US evaluation set for the navigation planner: 19,744 records, each pairing a single front-camera frame with the corresponding past trajectory, future ground-truth waypoints, and a natural-language driving objective.

Provenance

Every record here was drawn — uniformly at random — from the pool of US scenes that were withheld from every training stage of the planner:

• the base VLA pretraining mix,
• the reasoning supervised fine-tuning (SFT) stage, and
• the GRPO reinforcement-learning stage.

In other words: no record in this directory was ever seen by the models that will be evaluated on it. This is the canonical "honest" US eval set — use it to compare checkpoints without leakage concerns.

The set is US-only by construction.

Directory layout

PhysicalAI-US-Evaluation/
├── README.md                ← this file
├── dataset.jsonl            ← 19,744 records, one JSON object per line
└── camera/
    ├── camera_00.zip        ← original packed frames (≤3 GB cap, ZIP_STORED)
    └── camera_01.zip

Image format

• Codec: PNG (cv2.IMWRITE_PNG_COMPRESSION=3)
• Resolution: 640 × 360
• Camera: camera_front_wide_120fov (front-facing, ~120° HFOV)
• Path layout: {chunk_name}/{scene_id}/{timestamp_us}.png, relative to this directory after unzipping the camera files.
• Frame selection: the frame whose timestamp is closest to clip_start_us + timestamp_us for the given scene.

Resolving a record's frame is therefore:

frame_path = root / record["chunk_name"] / record["scene_id"] / f'{record["timestamp_us"]}.png'

JSONL schema

Each line in dataset.jsonl is a single JSON object. Fields fall into two groups.

Identity / indexing (always present)

Field	Type	Meaning
shard_id	str	Source shard the row was drawn from (e.g. "shard_00020").
chunk_name	str	Chunk directory the frame lives in (e.g. "chunk_1580").
scene_id	str	UUID of the scene clip.
timestamp_us	int	Offset into the clip, in microseconds. Used to address the frame.

Trajectory + task (always present)

Field	Type	Meaning
egomotion	list[[x, y, θ]]	Past trajectory, ego-frame, 0.25 s spacing, ending at the present anchor [0, 0, 0]. Length 9 (= 2 s of history + anchor).
ground_truth_waypoints	list[[x, y, θ]]	Future trajectory, ego-frame, 0.25 s spacing. Length 24 (= 6 s of lookahead).
objective	str	Natural-language driving intent for the next few seconds (e.g. "Drive straight along the multi-lane road").
difficulty	"easy" | "hard"	See Difficulty split below.

Coordinates are right-handed ego-frame, units meters / radians, with +x forward and +y left.

Difficulty split

Difficulty	Count	Meaning
hard	2,000	Scenes selected up-front as harder, non-trivial driving situations. Use these when you want a quality bar that exercises reasoning behavior.
easy	17,744	Uniformly random scenes from the held-out US pool, gated only on "the frame renders and at least one ground-truth waypoint is visible." A broad coverage sample.

The --hard flag in the evaluation harness keeps only the 2k hard rows; without it you evaluate on the full 19,744.

Using this dataset

This directory is the default data source for the US run of the ADE evaluation harness:

python GRPO-alignment/evaluation/AverageDisplacementError.py \
    --country US \
    --models <model-subdir-on-HF>

The harness reads dataset.jsonl, resolves each frame via the path layout above, and reports an Average Displacement Error leaderboard. Add --hard to restrict the evaluation to the 2,000 hard rows.

For full provenance details (random seed, source paths, render-gating reject reasons) see build_manifest.json and build_dataset.py.