NVIDIA
Alpamayo 1.5 / 2 Super
What Alpamayo 1.5 / 2 Super’s own documents say it is built from — every method with the sentence that describes it.
- Parameters
- 10B
- Attention
- Cosmos-Reason2-8B backbone; reasoning text + driving trajectory
- Input
- text+image -> action
- Licence
- OpenMDW-1.1
- Released
- 2026-03-03
Against the consensus recipe
Its documents state 1 of the 18 methods the field agrees on.
Data curation
- Benchmark decontaminationnot in its documents
- Synthetic data generationnot in its documents
- Corpus deduplicationnot in its documents
Training objective
- Multi-Token Predictionnot in its documents
- Knowledge insulationused
- Quantization-aware distillationnot in its documents
Model architecture
- Mixture of expertsnot in its documents
- Multi-Token Predictionnot in its documents
- Grouped-query attentionnot in its documents
Optimization
- Muonnot in its documents
- Expert parallelismnot in its documents
- Quantization-Aware Trainingnot in its documents
Post-training
- Supervised fine-tuningnot in its documents
- Asynchronous reinforcement learningnot in its documents
- Multi-Teacher On-Policy Distillationnot in its documents
Inference & serving
- Thinking modenot in its documents
- Multi-Token Predictionnot in its documents
- FP8not in its documents
prediction head 1
“featuring a diffusion-based trajectory decoder”
multimodal architecture 3
“we adopt a separate action-expert to decode actions via flow matching”
“NVIDIA’s open 10B-parameter chain-of-thought reasoning VLA model”
“E2E approaches map raw sensor inputs directly to vehicle motion through jointly trained neural networks.”
1 more mentioned or evaluated
multimodal architecture “the triplane-based multi-camera tokenizer from Ivanovic et al. (2025) achieves nearly identical minADE6 values as the baseline”
language modelling objective 1
“we apply the cross-entropy loss over both the reasoning tokens and the discrete trajectory tokens”
auxiliary loss 2
“During training, we apply a stop-gradient to the KV-cache produced by the VLM to prevent gradients from the expert back-propagating into…”
“The KL regularization term with coefficient penalizes deviations from the reference policy (typically the SFT model)”
1 more mentioned or evaluated
language modelling objective “a baseline model trained to auto-regressively predict 6 discrete trajectory tokens”
filed at the root 1
“We develop a structured Chain of Causation (CoC) labeling framework that produces decision-grounded, causally-linked reasoning traces”
data filtering 7
“We therefore prioritize these high-disagreement samples to construct a focused post-training dataset, while mixing in a similar proportion…”
“Each labeled instance first undergoes a quality check performed by a different annotator.”
“Selecting keyframes for CoC annotation is therefore critical to maximizing the clarity of decision causality.”
“For proactive scenarios, we annotate a keyframe range: a time window during which the ego actively evaluates or prepares for a potential…”
“implement corresponding rule-based detectors to infer these meta actions at the frame level.”
“We employ rule-based methods to identify clips corresponding to each scenario and balance the number of clips per scenario to ensure…”
“apply a safety exclusion filter to remove invalid data with illegal or unsafe driving behavior”
synthetic data 4
“We first define a comprehensive set of high-level driving decisions that directly correspond to low-level ego trajectories. Each reasoning…”
“VLMs are prompted to use the 2-second historical video when identifying critical components.”
“Automated (VLM-based auto-labeling and heuristic rules)”
“we employ state-of-the-art VLMs such as GPT-5 (OpenAI, 2025) to perform offline auto-labeling through a multi-step reasoning process.”
tokenization 2
“AR1 adopts single-image tokenization by default”
“During training, we inject the action modality to the VLM through discrete tokens”
filed at the root 2
“We adopt a multi-stage training strategy that builds upon the Cosmos-Reason VLM backbone”
“we adopt a three-stage training strategy to transform the VLM into a reasoning-capable autonomous driving policy.”
supervised fine-tuning 1
“We perform SFT on the CoC dataset to teach the model to generate reasoning traces through imitation”
reinforcement learning algorithm 3
“we introduce an RL-based post-training framework shown in Fig. 6 that optimizes three complementary reward signals”
“we apply RL-based post-training to simultaneously improve reasoning quality, reasoning-action consistency, and trajectory quality”
“We adopt GRPO (Shao et al., 2024) as our alignment algorithm.”
reward modelling 4
“we introduce a reasoning reward derived from LRM feedback”
“we introduce a CoC–action consistency reward that measures behavioral alignment between the generated reasoning trace and the…”
“we introduce a safety reward that penalizes unsafe or physically implausible trajectories during post-training”
“The reward combines three terms:”
mid-training & continual pretraining 1
“Alpamayo-R1 adopts Cosmos-Reason as its VLM backbone, specifically post-trained on Physical AI data including driving scenarios.”
filed at the root 1
“By reformulating these aspects as a set of True/False questions, the evaluation process becomes more interpretable and better aligned with…”
judge 1
“we employ large reasoning models (LRMs) as automatic evaluators to provide scalable, high-quality feedback on reasoning quality.”
human & real-world evaluation 3
“Closed-Loop Evaluation using AlpaSim on 913 scenarios”
“we adopt a hybrid evaluation strategy that combines human verification with LLM-based auto-evaluation.”
“Proprietary autonomous driving test datasets, closed-loop simulation, on-vehicle road tests.”
other 3
filed at the root 3
“For reactive scenarios, a keyframe is typically chosen by applying a short temporal buffer (approximately 0.5 seconds) before the ego…”
“human annotators are required to complete a two-stage procedure designed to produce concise and causally grounded CoC write-ups.”
“we adopt an action representation governed by unicycle dynamics that leads to better closed-loop performance”
unfiled 19
“Both models generate Chain-of-Causation reasoning traces alongside trajectory predictions.”
“We employ a diffusion-based action-expert trajectory decoder built on flow matching”
“we develop a hybrid labeling procedure that combines human labeling and auto-labeling.”
“constructed through a hybrid labeling pipeline that combines large-scale auto-labeling with humans in the loop”
“During inference, we set , i.e., 5 steps, in flow matching to reduce latency with negligible performance degradation.”
“we define a closed set of high-level driving decisions as in Tab. 1. Each clip is annotated with at most one longitudinal and one lateral…”
“We train the action-expert using a vanilla conditional flow matching loss”
“adopting a continuous action representation governed by unicycle dynamics with flow matching”
“During inference, starting with , we perform denoising through Euler integration:”
“In practice, we adopt the Gaussian conditional optimal transport (OT) path”
“supported by a hybrid human-in-the-loop and auto-labeling pipeline for scalable high-quality data generation”
“combining discrete trajectory tokens learned within the VLM with an action-expert that decodes the same trajectories into continuous…”
“a keyframe is typically chosen by applying a short temporal buffer (approximately 0.5 seconds) before the ego vehicle initiates a behavior…”
“compare it against the meta-actions derived from the predicted trajectory using rule-based matching.”
“the ground-truth control sequence is derived from through a least-squares formulation with Tikhonov regularization to attenuate…”
4 more mentioned or evaluated
“Flex (Yang et al., 2025) is able to achieve more drastic improvements”
“we observe consistent improvements in open-loop performance as model size increases”
“We train the 0.5B model on varying amounts of data: 100k, 200k, 500k, 1M, and 2M video segments, keeping the total number of training…”
“Following the V-model methodology, iterative testing and validation at both unit and system levels are essential”
Sources
The curated source list for this model. "not fetched" means the URL is recorded but its text was not read in the current run.
| Source | Kind | Publisher | Status |
|---|---|---|---|
| huggingface.co/nvidia/Alpamayo-1.5-10B | model card | NVIDIA | read |
| arxiv.org/abs/2511.00088 | technical report | NVIDIA | read |