Add links to paper page, Github repository and project page
Browse filesThis PR adds links to:
- The paper at https://huggingface.co/papers/2506.16655
- The Github repository
- A project page at https://archgw.com/
README.md
CHANGED
@@ -1,14 +1,13 @@
|
|
1 |
---
|
2 |
-
license: other
|
3 |
-
license_name: katanemo-research
|
4 |
-
license_link: >-
|
5 |
-
https://huggingface.co/katanemo/Arch-Router-1.5B.gguf/blob/main/LICENSE
|
6 |
base_model:
|
7 |
- Qwen/Qwen2.5-1.5B-Instruct
|
8 |
language:
|
9 |
- en
|
10 |
-
pipeline_tag: text-generation
|
11 |
library_name: transformers
|
|
|
|
|
|
|
|
|
12 |
---
|
13 |
|
14 |
# katanemo/Arch-Router-1.5B
|
@@ -19,7 +18,7 @@ With the rapid proliferation of large language models (LLMs) -- each optimized f
|
|
19 |
|
20 |
We introduce a preference-aligned routing framework that guides model selection by matching queries to user-defined domains (e.g., travel) or action types (e.g., image editing) -- offering a practical mechanism to encode preferences in routing decisions. Specifically, we introduce Arch-Router, a compact 1.5B model that learns to map queries to domain-action preferences for model routing decisions. Experiments on conversational datasets demonstrate that our approach achieves state-of-the-art (SOTA) results in matching queries with human preferences, outperforming top proprietary models.
|
21 |
|
22 |
-
This model is described in the paper: https://arxiv.org/abs/2506.16655, and powers [Arch](https://github.com/katanemo/arch) the open-source AI-native proxy for agents to enable preference-based routing in a seamless way.
|
23 |
|
24 |
### How It Works
|
25 |
|
@@ -62,11 +61,15 @@ TASK_INSTRUCTION = """
|
|
62 |
You are a helpful assistant designed to find the best suited route.
|
63 |
You are provided with route description within <routes></routes> XML tags:
|
64 |
<routes>
|
65 |
-
|
|
|
|
|
66 |
</routes>
|
67 |
|
68 |
<conversation>
|
69 |
-
|
|
|
|
|
70 |
</conversation>
|
71 |
"""
|
72 |
|
|
|
1 |
---
|
|
|
|
|
|
|
|
|
2 |
base_model:
|
3 |
- Qwen/Qwen2.5-1.5B-Instruct
|
4 |
language:
|
5 |
- en
|
|
|
6 |
library_name: transformers
|
7 |
+
license: other
|
8 |
+
license_name: katanemo-research
|
9 |
+
license_link: https://huggingface.co/katanemo/Arch-Router-1.5B.gguf/blob/main/LICENSE
|
10 |
+
pipeline_tag: text-generation
|
11 |
---
|
12 |
|
13 |
# katanemo/Arch-Router-1.5B
|
|
|
18 |
|
19 |
We introduce a preference-aligned routing framework that guides model selection by matching queries to user-defined domains (e.g., travel) or action types (e.g., image editing) -- offering a practical mechanism to encode preferences in routing decisions. Specifically, we introduce Arch-Router, a compact 1.5B model that learns to map queries to domain-action preferences for model routing decisions. Experiments on conversational datasets demonstrate that our approach achieves state-of-the-art (SOTA) results in matching queries with human preferences, outperforming top proprietary models.
|
20 |
|
21 |
+
This model is described in the paper: https://arxiv.org/abs/2506.16655, and powers [Arch](https://github.com/katanemo/arch) the open-source AI-native proxy for agents to enable preference-based routing in a seamless way. The project page can be found at https://archgw.com/.
|
22 |
|
23 |
### How It Works
|
24 |
|
|
|
61 |
You are a helpful assistant designed to find the best suited route.
|
62 |
You are provided with route description within <routes></routes> XML tags:
|
63 |
<routes>
|
64 |
+
|
65 |
+
{routes}
|
66 |
+
|
67 |
</routes>
|
68 |
|
69 |
<conversation>
|
70 |
+
|
71 |
+
{conversation}
|
72 |
+
|
73 |
</conversation>
|
74 |
"""
|
75 |
|