BADAS 2.0 – The World's Best Incident Prediction Model

02 – The Open Challenge

0.994 AP vs 0.940. Same Data.
Different Architecture.

We fine-tuned NVIDIA's COSMOS-Reason2-2B on the exact same 2M training clips. BADAS 2.0 achieves 99.4% AP. COSMOS achieves 94.0%. At 91x fewer parameters, our smallest model still beats their largest. We publish our benchmarks and invite direct comparison.

Metric	BADAS 2.0	COSMOS-BADAS
Average Precision	99.4%	94.0%
Early Warning Recall	91.3%	48.3%
Architecture	V-JEPA2 (Attention)	Autoregressive
Smallest Model	22M params	2B params (cloud only)
Training Data	2M real-world clips	2M real-world clips (same)
Explainability	Native attention maps	None

COSMOS-BADAS = NVIDIA COSMOS-Reason2-2B fine-tuned on the same 2M Nexar training clips used by BADAS 2.0.

The Efficiency Gap: Fine-tuning improved COSMOS by +18.5 pp AP – but at 2B parameters, it still sits 5.4 pp below BADAS 2.0 (94.0% vs 99.4%). BADAS 2.0 Flash Lite (22M) outperforms COSMOS-BADAS (2,000M) by +4.4 pp AP while being 91x smaller.

Early Warning Recall

48.3%

COSMOS-BADAS

91.3%

BADAS 2.0

In collision anticipation, late is too late.

The Efficiency Gap

Parameters (M) → log scale

100.0%

99.0%

96.0%

93.0%

90.0%

BADAS 2.0 Flash Lite (22M) outperforms COSMOS-BADAS (2,000M) – 91× smaller.

BADAS 2.0 — 300M, 99.4% AP

BADAS 1.0 — 300M, 96.0% AP

BADAS 2.0 Flash — 86M, 99.0% AP

BADAS 2.0 Flash Lite — 22M, 98.4% AP

COSMOS-BADAS — 2B, 94.0% AP

COSMOS-Reason2 — 2B, 75.6% AP

03 – The Proof

We Published Everything. Your Move.

Long-Tail Benchmark: Per-Category Breakdown

AUC and AP across 10 scenario groups (888 clips, sliding window). BADAS 2.0 leads in every category. Best per row in bold.

Group		BADAS 2.0	BADAS 2.0Flash	BADAS 2.0Flash Lite	BADAS 1.0	BADAS Open	COSMOS-Reason2Fine-tuned	Gemini 2.5 ProFine-tuned	Qwen3-VL-2B
Animal	AUC	96.4%	94.8%	92.3%	94.8%	88.1%	81.6%	79.9%	75.9%
Animal	AP	99.1%	98.8%	98.1%	98.7%	95.7%	95.2%	93.9%	93.2%
Pedestrian	AUC	99.8%	99.6%	99.1%	99.1%	84.4%	93.6%	77.4%	67.4%
Pedestrian	AP	99.9%	99.7%	99.4%	99.4%	87.9%	95.6%	77.5%	75.5%
Intersection	AUC	100.0%	100.0%	99.8%	98.8%	95.8%	97.9%	85.0%	61.7%
Intersection	AP	100.0%	100.0%	99.7%	98.4%	93.7%	96.9%	73.7%	45.3%
Overtaking	AUC	100.0%	100.0%	99.7%	97.4%	92.7%	97.5%	83.2%	82.6%
Overtaking	AP	100.0%	99.9%	99.4%	96.1%	85.3%	94.8%	60.8%	68.0%
Snow	AUC	100.0%	100.0%	99.9%	99.6%	93.2%	97.5%	95.3%	80.4%
Snow	AP	100.0%	100.0%	99.9%	99.5%	94.0%	97.3%	93.6%	77.9%
Infrastructure	AUC	100.0%	99.4%	98.4%	86.9%	84.0%	97.4%	90.8%	58.4%
Infrastructure	AP	100.0%	99.6%	98.7%	91.5%	88.0%	98.1%	92.2%	62.5%
Motorcyclist	AUC	99.8%	99.8%	100.0%	99.8%	96.6%	98.9%	80.1%	72.5%
Motorcyclist	AP	99.9%	99.9%	100.0%	99.9%	96.9%	99.1%	76.6%	76.1%
Cyclist	AUC	100.0%	98.7%	99.4%	98.6%	93.1%	94.0%	82.9%	64.8%
Cyclist	AP	100.0%	99.1%	99.5%	98.7%	94.2%	95.3%	81.2%	59.8%
Rain	AUC	100.0%	100.0%	100.0%	97.5%	96.6%	99.6%	82.2%	81.5%
Rain	AP	100.0%	100.0%	100.0%	98.0%	95.9%	99.4%	64.5%	66.2%
Fog	AUC	100.0%	99.8%	99.8%	100.0%	99.4%	98.2%	81.6%	82.5%
Fog	AP	100.0%	99.5%	99.5%	100.0%	98.7%	95.9%	60.5%	76.0%
OVERALL	AUC	99.3%	98.9%	98.1%	94.9%	82.3%	92.6%	83.3%	67.8%
OVERALL	AP	99.4%	99.0%	98.4%	96.0%	84.5%	94.1%	79.5%	67.2%

Nexar Kaggle Benchmark

Single-window mean AP over three lead-time thresholds (1,344 clips). BADAS 2.0 improves mAP from 92.5% to 94.0% while cutting the false positive rate by 74%.

Model	AP @0.5s	AP @1.0s	AP @1.5s	mAP	FPR	Params
BADAS 2.0	94.3%	95.7%	92.1%	94.0%	4.6%	300M
BADAS 2.0 Flash	94.5%	96.2%	91.5%	94.1%	9.7%	86M
BADAS 2.0 Flash Lite	94.6%	94.7%	90.7%	93.3%	12.2%	22M
BADAS 1.0	93.5%	93.6%	90.4%	92.5%	10.9%	300M
COSMOS-BADAS	90.4%	88.9%	87.5%	88.9%	–	2B

Reading this table: AP @0.5s / @1.0s / @1.5s – Average Precision measured at three different lead times before the collision. Higher = better detection at that warning horizon. mAP – Mean Average Precision, the average of the three AP scores above. FPR – False Positive Rate. Lower = fewer false alarms. Params – Model size in parameters.

74% Fewer False Alarms

On the internal test set, BADAS 2.0 cuts the false positive rate from 17.7% (v1.0) to 4.6% – a 74% reduction with no loss of recall.

4.6%

FPR – BADAS 2.0

mAP 94.0% · 300M params · 34ms

9.7%

FPR – BADAS 2.0 Flash

mAP 94.1% · 86M params · 4.8ms

10.9%

FPR – BADAS 1.0

mAP 92.5% · 300M params · 2,500ms

Early Warning Recall (Long-Tail Benchmark)

Fraction of collision events detected before they occur (888 clips, 10 scenario groups, threshold 0.75).

BADAS 2.0

91.3%

F1 96.4%

BADAS 2.0 Flash

89.9%

F1 93.8%

BADAS 1.0

85.5%

F1 87.6%

External Benchmarks (Sliding Window)

AUC and AP on three public academic benchmarks using ego-centric re-annotation. Best per column in bold.

Model	DAD AUC	DAD AP	DoTA AUC	DoTA AP	DADA AUC	DADA AP
BADAS 2.0	99.3%	92.2%	99.1%	99.9%	99.1%	99.6%
BADAS 2.0 Flash	98.7%	84.9%	98.5%	99.8%	99.0%	99.5%
BADAS 2.0 Flash Lite	98.2%	87.0%	98.5%	99.8%	98.1%	99.2%
BADAS 1.0	99.0%	94.0%	72.0%	95.0%	87.0%	90.0%
COSMOS-BADAS	94.4%	60.2%	98.3%	99.8%	95.9%	97.8%
Qwen3-VL-2B	75.4%	14.1%	70.9%	95.1%	80.5%	88.6%

Reading this table: AUC – Area Under the ROC Curve. Measures how well the model separates collisions from safe driving. 100% = perfect. AP – Average Precision. DAD, DoTA, DADA-2000 – Three public academic collision anticipation benchmarks with re-annotated ego-centric protocol.

How Confidence Evolves Over Time

Average collision probability over normalized pre-event time (0% = start, 100% = event). Each clip's timeline is scaled independently, so clips of different lengths are comparable. Positive clips only. BADAS models ramp up sharply; competitors stay flat.

Reading this chart: For every positive clip, each model's prediction timeline is normalized so 0% = first prediction and 100% = labeled event. Predictions are binned into 10 equal intervals then averaged across clips. Curves are baseline-normalized per model so the y-axis shows each model's rise above its own floor. A steep ramp means confidence increases sharply as the event approaches; a flat line means the model outputs a near-constant score regardless of proximity to collision.

04 – See What the Model Sees

See What the Model Sees.
Know Why It Acts.

For the first time, a collision anticipation system explains itself. Attention heatmaps show where the model looks. BADAS-Reason tells you what to do and why. OEMs get integration confidence. Insurers get audit trails. Fleet operators get actionable alerts.

Explainability

Attention heatmaps reveal exactly what the model sees and focuses on during risk scenarios. Not a black box.

Attention heatmap – spatial focus shifts to the approaching vehicle

Actionability

Predicts the right action to take and explains why in natural language. "Brake immediately – a dark vehicle is crossing the intersection from the left directly into the ego vehicle's path."

BADAS-Reason – natural language action recommendation with reasoning

05 – From Cloud to Edge

Three Models. From Cloud to Edge.
GPU and CPU.

Cloud analytics teams run the full 300M model. Edge ADAS integrators deploy BADAS 2.0 Flash. IoT manufacturers ship BADAS 2.0 Flash Lite. Same architecture, same training, same world model – scaled to your needs.

BADAS 2.0

300M parameters

SOTA performance across all metrics
Best mTTA and early warning recall
Expert in rare long-tail cases
34ms on A100 · 41ms on Jetson Thor

BADAS 2.0 Flash

86M parameters

End-device-optimized model
Expert in false alarm prevention
Outperforms BADAS 1.0 on every metric
4.8ms on A100 · 12.5ms on Jetson Thor

BADAS 2.0 Flash Lite

22M parameters

Ultra-light model for IoT devices
Optimized for GPU and CPU inference
Rivals BADAS 1.0 at 14x fewer params
2.8ms on A100 · 5.9ms on Jetson Thor

Performance vs Latency

Model	Params	AP	A100 (FP16)	Jetson Thor (TensorRT)
BADAS 2.0	300M	99.4%	34ms	41ms
BADAS 2.0 Flash	86M	99.0%	4.8ms	12.5ms
BADAS 2.0 Flash Lite	22M	98.4%	2.8ms	5.9ms

BADAS 2.0 Flash Lite loses only 1% AP while running 12x faster than the full model on A100 – and 7x faster on Thor. Deploy anywhere from cloud to edge device.

06 – The Science

V-JEPA2 Validates the World Model Thesis.

BADAS 2.0 is built on V-JEPA2 – the architecture Yann LeCun proposed as the foundation for world models and physical AI. Latent-space prediction optimizes for physical causality, not visual reconstruction. BADAS 2.0 is the proof that this thesis works for safety-critical applications.

350K Cameras

Real-world capture
100M+ miles/month

→

Nexar Atlas

GPS-validated pipeline
45 PB structured video

→

V-JEPA2

Self-supervised learning
Latent-space prediction

→

BADAS 2.0

94.0% mAP · 4.6% FPR
34ms · 91x smaller

BADAS 2.0 fine-tunes a V-JEPA2 ViT-L backbone (300M parameters, 24 transformer layers) end-to-end on 16-frame clips at 256×256 resolution and 8 fps. A future-prediction branch estimates the scene 1 second ahead and concatenates it with the current clip, giving the prediction head access to both present evidence and near-future dynamics. Domain-specific SSL pre-training on 2.25M unlabeled Nexar edge device clips is the critical enabler for the distilled edge variants.

Why V-JEPA2 matters: V-JEPA2 learns by predicting the latent-space representation of future video frames rather than reconstructing pixels. Pixel reconstruction optimizes for visual fidelity. Latent-space prediction optimizes for physical causality. For collision anticipation, you need a model that understands what will happen – not just what is happening.

~200,000 Labeled Videos. Zero Synthetic Data.

Most collision anticipation models are trained on synthetic data or small academic datasets. BADAS 2.0 is trained exclusively on real-world edge device footage from Nexar's network – the largest ego-centric driving dataset ever assembled for this task.

BADAS 2.0 is trained on ~200,000 labeled videos (~2M windowed clips) – a 5x expansion over v1.0. The corpus is assembled through intelligent data mining: BADAS 1.0 runs as an active oracle over millions of unlabeled Nexar drives, surfacing high-risk clips for human review.

The result: 99.4% AP at 4.6% FPR – a 58% reduction in false alarms over v1.0 on the sliding-window benchmark, with gains across all subgroups including the hardest long-tail categories.

Same Architecture. Better Data. Much Better Results.

1.5K

BADAS Open

~40K

BADAS 1.0

~200K

BADAS 2.0

Long Tail of Driving

Excels on rare, edge-case scenarios – animals, fog, snow, motorcyclists, infrastructure failures. 99.4% AP across all 10 long-tail categories where competitors collapse.

Physics, Not Pattern Matching

V-JEPA2 predicts latent-space representations of future frames. This optimizes for physical causality – what will happen – not visual similarity to training data.

World Model

Beyond Driving

Works on situations that have nothing to do with driving on the road directly – drones flying, vacuums cleaning, a forklift at work. This is because the model went beyond road rules. It learned physics.

BADAS 2.0 is not a driving model. It is a world model that happens to be deployed on roads.

Per-Category Dominance

BADAS 2.0 vs COSMOS across all 10 long-tail categories. BADAS leads in every single one.

BADAS 2.0

COSMOS-BADAS

COSMOS-Reason2

Models don't emerge from abstractions alone. They come from sustained exposure to reality.

– Yann LeCun, Turing Award Winner, Nexar Board Member

07 – Deployment

Where BADAS Deploys Today

AV Program Development

Ground truth for training, validating, and benchmarking autonomous systems. Access the world's largest library of outcome-verified edge cases.

ADAS Supplier Integration

API access to collision prediction as a feature layer. Ship BADAS as a premium safety tier on Snapdragon Ride or proprietary ADAS platforms.

Fleet Safety

Real-time collision risk scoring for commercial vehicle operations. Move from reactive incident management to predictive intervention.

Insurance Underwriting

BADAS risk scores as underwriting inputs. Behavior-based pricing backed by production-grade collision anticipation AI.

08 – FAQ

Frequently Asked Questions

BADAS stands for Beyond ADAS (Advanced Driver Assistance Systems). It is Nexar's collision anticipation system, now in its second generation. BADAS 2.0 fine-tunes V-JEPA2 on ~200,000 labeled edge device videos (~2M windowed clips) and achieves state-of-the-art accuracy across all public collision anticipation benchmarks.

BADAS 2.0 was evaluated on the Nexar Kaggle competition (1,344 clips, single window), a new 10-group long-tail benchmark (888 clips covering animal, pedestrian, cyclist, fog, rain, snow, intersection, infrastructure, passing/overtaking, and motorcyclist scenarios), and three public external benchmarks: DAD, DoTA, and DADA-2000 using ego-centric re-annotation and sliding-window evaluation.

The paper compares five BADAS variants (2.0, 1.0, 2.0 Flash, 2.0 Flash Lite, Open) against four VLM baselines: Gemini-BADAS (Gemini 2.5 Pro fine-tuned on BADAS data), COSMOS-BADAS (NVIDIA COSMOS-Reason2-2B fine-tuned), vanilla Gemini 2.5 Pro, and Qwen3-VL-2B. Even after fine-tuning on the same data, autoregressive VLMs remain significantly below the BADAS family on the long-tail benchmark.

BADAS 2.0 fine-tunes V-JEPA2 (ViT-L, 300M parameters) end-to-end on edge device video. A future-prediction branch estimates the scene 1 second ahead, giving the classifier access to both present and anticipated dynamics. The distilled variants – BADAS 2.0 Flash at 86M (4x compression) and BADAS 2.0 Flash Lite at 22M (14x compression) – use domain-specific SSL pre-training followed by knowledge distillation to achieve near-parity accuracy at 7–12x faster inference.

BADAS is available via a public API that lets you upload video, run predictions, and export results – including prediction overlays and attention heatmaps. Try the web-based playground at the top of this page for direct browser access. Enterprise partners can request full API access for integration into AV programs, fleets, or ADAS platforms.