Add LAIR overview and research resource links
Browse files
README.md
CHANGED
|
@@ -7,4 +7,39 @@ sdk: static
|
|
| 7 |
pinned: false
|
| 8 |
---
|
| 9 |
|
| 10 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 7 |
pinned: false
|
| 8 |
---
|
| 9 |
|
| 10 |
+
|
| 11 |
+
|
| 12 |
+
## Lehigh AI Research Lab
|
| 13 |
+
|
| 14 |
+
We are **LAIR at Lehigh University**, led by [Prof. Lichao Sun](https://lichao-sun.github.io/). We study how to make AI systems more reliable, robust, efficient, and interpretable, and build tools for AI-assisted scientific research.
|
| 15 |
+
|
| 16 |
+
[Lab website](https://openlair.github.io/) · [GitHub](https://github.com/OpenLAIR) · [Publications](https://openlair.github.io/publications/) · [People](https://openlair.github.io/people/)
|
| 17 |
+
|
| 18 |
+
### Research areas
|
| 19 |
+
|
| 20 |
+
Trustworthy AI · Medical AI · Generative AI and large language models · Federated learning · Reinforcement learning
|
| 21 |
+
|
| 22 |
+
Our recent projects explore scientific agents, looped decision models, and the evaluation of human and AI-generated peer reviews.
|
| 23 |
+
|
| 24 |
+
### Models and datasets
|
| 25 |
+
|
| 26 |
+
The following resources accompany our research projects. They are currently hosted under the **songdj** account; the links below point to their original model and dataset cards.
|
| 27 |
+
|
| 28 |
+
| Project | Resources | What you will find |
|
| 29 |
+
|---|---|---|
|
| 30 |
+
| **LEV** | [Lev-1.4B](https://huggingface.co/songdj/Lev-1.4B) · [Lev-2.6B](https://huggingface.co/songdj/Lev-2.6B) · [Lev-3B](https://huggingface.co/songdj/Lev-3B) | Looped decision models with multiple inference depths. |
|
| 31 |
+
| **LEV evaluation** | [LevBench](https://huggingface.co/datasets/songdj/LevBench) | Development and locked-test decisions across five tiers; a rights-filtered subset. |
|
| 32 |
+
| **LEV training** | [Lev-train-v2](https://huggingface.co/datasets/songdj/Lev-train-v2) | Training mixtures and recipes; a rights-filtered subset. |
|
| 33 |
+
| **LEV results** | [Lev-results](https://huggingface.co/datasets/songdj/Lev-results) | Per-depth outputs, calibration, and latency evidence, without question text. |
|
| 34 |
+
| **PremiseCheck** | [ReviewFact](https://huggingface.co/datasets/songdj/ReviewFact) | A benchmark for factual errors in peer-review comments. For evaluation, not model training. |
|
| 35 |
+
|
| 36 |
+
Explore the [LEV collection](https://huggingface.co/collections/songdj/lev), [LEV code](https://github.com/OpenLAIR/lev), and [PremiseCheck code](https://github.com/OpenLAIR/PremiseCheck). Please consult each resource card for its intended use, limitations, and license terms.
|
| 37 |
+
|
| 38 |
+
### Research tools and projects
|
| 39 |
+
|
| 40 |
+
- **[Dr. Claw](https://github.com/OpenLAIR/dr-claw)** — An AI research workspace for organizing agent-assisted scientific workflows.
|
| 41 |
+
- **[OpenSkill](https://github.com/OpenLAIR/OpenSkill)** — Research on agents that develop skills and verification signals through self-evolution.
|
| 42 |
+
- **[ClawChain](https://github.com/OpenLAIR/ClawChain)** — Runtime monitoring, recovery, and evidence tracking for AI agents.
|
| 43 |
+
- **[Awesome Vibe Researching](https://github.com/OpenLAIR/awesome-vibe-researching)** — A curated collection of resources for AI-driven research automation.
|
| 44 |
+
|
| 45 |
+
For research updates, collaboration, and open positions, visit our [lab website](https://openlair.github.io/).
|