Upload Leaderboard.vue
Browse files
src/views/Leaderboard.vue
CHANGED
|
@@ -51,10 +51,6 @@ const lastSelectedDataNameChart = ref('')
|
|
| 51 |
|
| 52 |
// header markdown 内容
|
| 53 |
const headerMarkdown = ref(`
|
| 54 |
-
<p align="center">
|
| 55 |
-
🏆 <a href="https://huggingface.co/spaces/TeleAI-AI-Flow/InformationCapacityLeaderboard"> Leaderboard</a>    |   
|
| 56 |
-
🖥️ <a href="https://github.com/TeleAI-AI-Flow/InformationCapacity">GitHub</a>    |    🤗 <a href="https://huggingface.co/datasets/TeleAI-AI-Flow/InformationCapacity">Hugging Face</a>   |    📑  <a href="https://www.arxiv.org/abs/2511.08066">Paper</a>
|
| 57 |
-
</p>
|
| 58 |
|
| 59 |
**Information Capacity** evaluates an LLM's **efficiency** based on text compression performance relative to computational complexity, harnessing the inherent correlation between **compression** and **intelligence**.
|
| 60 |
Larger models can predict the next token more accurately, leading to higher compression gains but at increased computational costs.
|
|
@@ -62,7 +58,9 @@ Consequently, a series of models with varying sizes exhibits **consistent** info
|
|
| 62 |
It also facilitates dynamic routing of different-sized models for efficient handling of tasks with varying difficulties, which is especially relevant to the device-edge-cloud infrastructure detailed in the **AI Flow** framework.
|
| 63 |
With the rapid evolution of edge intelligence, we believe that this hierarchical network will replace the mainstream cloud-centric computing scheme in the near future.
|
| 64 |
|
|
|
|
| 65 |
If you want to add your evaluation results to the leaderboard, please submit a PR at [our GitHub repo](https://github.com/TeleAI-AI-Flow/InformationCapacity).
|
|
|
|
| 66 |
`)
|
| 67 |
|
| 68 |
const title = 'Information Capacity Leaderboard'
|
|
|
|
| 51 |
|
| 52 |
// header markdown 内容
|
| 53 |
const headerMarkdown = ref(`
|
|
|
|
|
|
|
|
|
|
|
|
|
| 54 |
|
| 55 |
**Information Capacity** evaluates an LLM's **efficiency** based on text compression performance relative to computational complexity, harnessing the inherent correlation between **compression** and **intelligence**.
|
| 56 |
Larger models can predict the next token more accurately, leading to higher compression gains but at increased computational costs.
|
|
|
|
| 58 |
It also facilitates dynamic routing of different-sized models for efficient handling of tasks with varying difficulties, which is especially relevant to the device-edge-cloud infrastructure detailed in the **AI Flow** framework.
|
| 59 |
With the rapid evolution of edge intelligence, we believe that this hierarchical network will replace the mainstream cloud-centric computing scheme in the near future.
|
| 60 |
|
| 61 |
+
|
| 62 |
If you want to add your evaluation results to the leaderboard, please submit a PR at [our GitHub repo](https://github.com/TeleAI-AI-Flow/InformationCapacity).
|
| 63 |
+
|
| 64 |
`)
|
| 65 |
|
| 66 |
const title = 'Information Capacity Leaderboard'
|