shubh-gupta-ai commited on
Commit
f9a136e
·
verified ·
1 Parent(s): 98d56d8

Update title, GitHub org, and description wording

Browse files
Files changed (1) hide show
  1. README.md +9 -8
README.md CHANGED
@@ -11,16 +11,17 @@ tags:
11
  - vllm
12
  ---
13
 
14
- # NoRD-Base: No Reasoning for Driving (SFT only)
15
 
16
- **CVPR 2026** | [arXiv](https://arxiv.org/abs/2602.21172) | [Project Page](https://nord-vla-ai.github.io/) | [GitHub](https://github.com/nord-vla-ai/nord)
17
 
18
  *Ishaan Rawal · Shubh Gupta · Yihan Hu · Wei Zhan*
19
 
20
  This is the **SFT-only ablation baseline** from the paper: [Qwen2.5-VL-3B](https://huggingface.co/Qwen/Qwen2.5-VL-3B-Instruct)
21
- supervised fine-tuned to directly predict driving trajectories as discrete tokens — no
22
- chain-of-thought reasoning, 3× fewer tokens than reasoning-based VLA baselines. This checkpoint
23
- is also the initialization used for the Dr. GRPO stage that produces the paper's main result.
 
24
 
25
  | Model | Training | NAVSIM PDMS |
26
  |---|---|---|
@@ -29,7 +30,7 @@ is also the initialization used for the Dr. GRPO stage that produces the paper's
29
 
30
  ## Usage
31
 
32
- Install the [`nord`](https://github.com/nord-vla-ai/nord) client and serve with vLLM:
33
 
34
  ```bash
35
  pip install "nord[serve]"
@@ -51,7 +52,7 @@ output = agent.predict(nord.NordInput(
51
  print(output.trajectory.shape) # (40, 3) — x, y, heading at 10 Hz, 4 seconds
52
  ```
53
 
54
- Full inference and NAVSIM evaluation instructions: [github.com/nord-vla-ai/nord](https://github.com/nord-vla-ai/nord).
55
 
56
  This repo bundles the K-Disc trajectory tokenizer vocab (`vocab.pkl`, 2048 clusters) used to
57
  decode the model's output tokens into `(x, y, heading)` trajectories.
@@ -71,4 +72,4 @@ decode the model's output tokens into `(x, y, heading)` trajectories.
71
 
72
  This checkpoint is released under [CC BY-NC-SA 4.0](https://creativecommons.org/licenses/by-nc-sa/4.0/)
73
  (non-commercial). The `nord` inference client code is separately licensed under Apache 2.0 —
74
- see [github.com/nord-vla-ai/nord](https://github.com/nord-vla-ai/nord).
 
11
  - vllm
12
  ---
13
 
14
+ # NoRD-Base: A Data-Efficient Vision-Language-Action Model that Drives without Reasoning (SFT only)
15
 
16
+ **CVPR 2026** | [arXiv](https://arxiv.org/abs/2602.21172) | [Project Page](https://nord-vla-ai.github.io/) | [GitHub](https://github.com/Applied-Intuition-Open-Source/nord)
17
 
18
  *Ishaan Rawal · Shubh Gupta · Yihan Hu · Wei Zhan*
19
 
20
  This is the **SFT-only ablation baseline** from the paper: [Qwen2.5-VL-3B](https://huggingface.co/Qwen/Qwen2.5-VL-3B-Instruct)
21
+ supervised fine-tuned to directly predict driving trajectories as discrete tokens without
22
+ chain-of-thought reasoning, using 3× fewer tokens than reasoning-based VLA baselines. This
23
+ checkpoint is also the initialization used for the Dr. GRPO stage that produces the paper's
24
+ main result.
25
 
26
  | Model | Training | NAVSIM PDMS |
27
  |---|---|---|
 
30
 
31
  ## Usage
32
 
33
+ Install the [`nord`](https://github.com/Applied-Intuition-Open-Source/nord) client and serve with vLLM:
34
 
35
  ```bash
36
  pip install "nord[serve]"
 
52
  print(output.trajectory.shape) # (40, 3) — x, y, heading at 10 Hz, 4 seconds
53
  ```
54
 
55
+ Full inference and NAVSIM evaluation instructions: [github.com/Applied-Intuition-Open-Source/nord](https://github.com/Applied-Intuition-Open-Source/nord).
56
 
57
  This repo bundles the K-Disc trajectory tokenizer vocab (`vocab.pkl`, 2048 clusters) used to
58
  decode the model's output tokens into `(x, y, heading)` trajectories.
 
72
 
73
  This checkpoint is released under [CC BY-NC-SA 4.0](https://creativecommons.org/licenses/by-nc-sa/4.0/)
74
  (non-commercial). The `nord` inference client code is separately licensed under Apache 2.0 —
75
+ see [github.com/Applied-Intuition-Open-Source/nord](https://github.com/Applied-Intuition-Open-Source/nord).