Add Büble

app/research/(projects)/bueble/page.tsx

@@ -0,0 +1,56 @@
+import React from "react";
+
+import BlockImage from "@/components/shared/BlockImage";
+import H2 from "@/components/shared/H2";
+import InlineLink from "@/components/shared/InlineLink";
+import P from "@/components/shared/P";
+import OL from "@/components/shared/UL";
+
+/**
+ * Set the title of the page here which is shown in the browser tab.
+ */
+export const metadata = {
+    title: "Büble-LM - Alan Akbik",
+};
+
+/**
+ * On this page you can insert the content for a project. You don't need to add the information you put inside the
+ * array inside @/content/Research.tsx. This will be automatically inserted for you. If you need certain HTML elements
+ * like a table or an unordered list, please check whether these components contain the functionality you need. If so use them
+ * instead of the standard HTML element.
+ *
+ * UnorderedList: <UL/>
+ * OrderedList: <OL/>
+ * Link: <InlineLink/>
+ * Paragraph: <P/>
+ * Table: <Table/>
+ * Image: <BlockImage/>
+ *
+ * These components can be imported from @/components/shared.
+ */
+export default function Page() {
+    return (
+        <>
+            <H2>Büble-LM</H2>
+            <P>
+                BübleLM is a state-of-the-art German language model based on Gemma-2B, adapted using trans-tokenization
+                with a custom German SentencePiece tokenizer.
+            </P>
+            <P>
+                Büble significantly outperforms other German LMs like Sauerkraut-2B and LLäMmlein-1B on most
+                benchmarks we tried. It was trained with a novel trans-tokenization approach by Pieter Delobelle
+                when he was a guest researcher at our chair!
+            </P>
+            <P>
+                More details on this model coming soon!
+            </P>
+            <H2>Getting Started</H2>
+            <P>
+                <OL>
+                    <li>Try out the <InlineLink href="https://huggingface.co/flair/bueble-lm-2b" target="_blank">model</InlineLink></li>
+                    <li>Check the <InlineLink href="https://huggingface.co/flair/bueble-lm-2b#performance" target="_blank">evaluation results</InlineLink></li>
+                </OL>
+            </P>
+        </>
+    );
+}

content/LatestNews.tsx

@@ -29,6 +29,14 @@ import type { PinnedMessage, TimelineEntryData } from "@/content/types";
  */
 
 const LatestNews: TimelineEntryData[] = [
+    {
+        date: "2024-12-05",
+        category: "New Model",
+        content: <>
+            Announcing 🐤 <InlineLink href="https://huggingface.co/flair/bueble-lm-2b" target="_blank">Büble-LM</InlineLink> 🐤,
+            our new state-of-the-art 2 billion parameter language model (LM) for German!
+        </>,
+    },
     {
         date: "2024-11-26",
         category: "New Project",
@@ -49,7 +57,7 @@ const LatestNews: TimelineEntryData[] = [
     {
         date: "2024-10-22",
         category: "Paper accepted",
-        conference: "BabyLM 2025",
+        conference: "BabyLM 2024",
         content: <>
             Our paper &quot;BabyHGRN: Exploring RNNs for Sample-Efficient Language Modeling&quot; accepted to BabyLM 2024!
         </>,

content/Research.tsx

@@ -36,6 +36,17 @@ const Research: ResearchProject[] = [
         categories: ["Featured", "Libraries"],
         uriComponent: "flair",
     },
+    {
+        title: "Büble-LM",
+        image: "/büble-lm.jpg",
+        imagePadding: 0.0,
+        imageFit: "contain",
+        introductoryText: <>
+            Büble-LM is our new state-of-the-art 2 billion parameter language model (LM) for German!
+        </>,
+        categories: ["Featured"],
+        uriComponent: "bueble",
+    },
     {
         title: "Zitatsuchmaschine",
         image: "/zitatsuchmaschine_logo.svg",
@@ -71,6 +82,17 @@ const Research: ResearchProject[] = [
         categories: ["Libraries", "Featured"],
         uriComponent: "transformer-ranker",
     },
+    {
+        title: "OpinionGPT",
+        image: "/opiniongpt.png",
+        imageFit: "contain",
+        imagePadding: 1,
+        introductoryText: <>
+            <InlineLink href="https://opiniongpt.informatik.hu-berlin.de/">OpinionGPT</InlineLink> is a ChatGPT-style model trained specifically to be biased and opinionated!
+        </>,
+        categories: ["Featured", "Applications"],
+        uriComponent: "opiniongpt",
+    },
     {
         title: "CleanCoNLL",
         image: "/CleanCoNLL_example_sentence.jpg",
@@ -84,17 +106,6 @@ const Research: ResearchProject[] = [
         categories: ["Featured", "Datasets"],
         uriComponent: "cleanconll",
     },
-    {
-        title: "OpinionGPT",
-        image: "/opiniongpt.png",
-        imageFit: "contain",
-        imagePadding: 1,
-        introductoryText: <>
-            <InlineLink href="https://opiniongpt.informatik.hu-berlin.de/">OpinionGPT</InlineLink> is a ChatGPT-style model trained specifically to be biased and opinionated!
-        </>,
-        categories: ["Featured", "Applications"],
-        uriComponent: "opiniongpt",
-    },
     {
         title: "LM Pub Quiz",
         image: "/BEAR-probe.svg",

content/types.d.ts

@@ -28,7 +28,7 @@ export type ConferenceAndYear = `${Conference} 20${number}${number}`
  *
  * If you need to add a new one you can extend this type.
  */
-export type NewsCategory = "New Paper" | "Paper accepted" | "Senior Area Chair" | "New Lab Member" | "New Research Grant" | "New Startup Grant" | "New Project"
+export type NewsCategory = "New Paper" | "Paper accepted" | "Senior Area Chair" | "New Lab Member" | "New Research Grant" | "New Startup Grant" | "New Project" | "New Model"
 
 export interface TimelineEntryData {
     date: IsoDate,