Update README.md: Fix image path and remove evaluation dataset section

README.md

@@ -21,7 +21,7 @@ tags:
 # **Align Your Flow: Scaling Continuous-Time Flow Map Distillation**
 **[Paper](https://arxiv.org/abs/2506.14603)** | **[Code](https://github.com/NVlabs/fastgen)** | **[Project Page](https://research.nvidia.com/labs/toronto-ai/AlignYourFlow/)**
 
-![](./assets/banner.jpg)
+![](assets/banner.jpg)
 
 # Introduction
 
@@ -160,18 +160,6 @@ We used a subset of the training data for testing and validation.
 - **Properties:** The testing dataset is approximately a 10% split of the training dataset. <br>
 
 
-### Evaluation Dataset:
-
-We evaluate our model on [GenEval](https://github.com/djghosh13/geneval) and [UniGenBench](https://github.com/CodeGoat24/UniGenBench) benchmarks. 
-
-- **Benchmark Score:** 0.62 on GenEval, 0.64 on UniGenBench <br>
-
-- **Data Collection Method by dataset:** Hybrid: Automated, Synthetic <br>
-
-- **Labeling Method by dataset:** Hybrid: Automated, Synthetic <br>
-
-- **Properties:** Both evaluation datasets consists of a set of text prompts. The generated images are evaluated using pretrained object detection models (GenEval) and pretrained large vision-language models VLMs (UniGenBench). <br>
-
 ## Key Considerations
 
 AYF models are trained as LoRA (Low-Rank Adaptation) extensions applied to existing text-to-image architectures. For this release, we fine-tuned [FLUX.1 \[dev\]](https://huggingface.co/black-forest-labs/FLUX.1-dev)  and [Qwen-Image](https://huggingface.co/Qwen/Qwen-Image) using lightweight LoRA adapters. This release includes only the LoRA adapters, not the underlying base models.