rwr

src/pages/index.mdx

@@ -22,7 +22,7 @@ import pipelinefig from "../../figs/progressor_method.png";
 import expfig from "../../figs/eval_tasks.png"
 import success_rate from "../../figs/success_rate.png"
 import train_reward from "../../figs/train_reward.png"
-
+import real_world_result from "../../figs/real_robot_eval.pdf"
 
 
 import transformer from "../assets/transformer.webp";
@@ -156,8 +156,17 @@ In our simulated experiments, we used benchmark tasks from the Meta-World enviro
 
 ###  Real-World Robotic Experiments
 
+#### Pretraining on Kitchen Dataset
+We randomly sample frame triplets triplet (<LaTeX inline formula="(\mathcal{o}_{i}, \mathcal{o}_{j}, \mathcal{o}_{g})" />) from the videos ensuring a maximal frame gap <LaTeX inline formula="|i − g| ≤ 2000".
+
+#### Real-World Few-Shot Offline Reinforcement Learning with Noisy Demonstrations
 
+We compare PROGRESSOR with R3M and VIP by freezing the pre-trained models and using them as reward prediction models to train RWR-ACT on downstream robotic learning tasks. 
 
+<Figure caption="real-world experiments">
+
+    <Image source={real_world_result} altText="real-world experiments" />
+</Figure>
 
 ## BibTeX citation