template is ready

papers.json

@@ -1,16 +1,9 @@
 [
   {
     "title": "MEGABYTE: Predicting Million-byte Sequences with Multiscale Transformers",
-    "authors": "Alice, Yu",
-    "year": 2024,
-    "link": "https://arxiv.org/abs/1234.5678",
-    "summary": "A survey on byte-based large language models."
-  },
-  {
-    "title": "Efficient Byte-Based Transformers",
-    "authors": "Charlie, Dana",
-    "year": 2023,
-    "link": "https://arxiv.org/abs/8765.4321",
-    "summary": "Exploration of efficient methods for byte-level transformers."
+    "date": "2023-05",
+    "link": "https://arxiv.org/pdf/2305.07185",
+    "conference": "NeurIPS'23",
+    "summary": "The paper introduces MEGABYTE, a multiscale Transformer architecture that segments sequences into patches, enabling efficient modeling of million-byte sequences with sub-quadratic self-attention, enhanced feedforward computation, and improved decoding parallelism, achieving competitive performance on tasks like long-context language modeling, image generation, and audio modeling."
   }
 ]

readme.py

@@ -5,17 +5,51 @@ def generate_markdown(json_file, output_file):
         papers = json.load(f)
 
     with open(output_file, 'w') as f:
-        # Write title and description
-        f.write("# Awesome Byte-Based Large Language Models\n")
-        f.write("A curated list of papers on byte-based large language models.\n\n")
+        f.write("<p align='center'>\n")
+        f.write("  <img src='assets/teaser.webp' alt='A teaser figure generated by DALL-E' width=70%>\n")
+        f.write("</p>\n\n")
+
+        # title
+        f.write("# Awesome Byte-Based Large Language Models\n\n")
+
+        # intro
+        f.write("## Introduction\n\n")
+
+        f.write(
+            "Recently, there has been a growing interest in studying byte-based large language models (LLMs). These models eliminate the need for tokenization and operate directly on raw bytes—the universal format of the digital world. Byte-based models offer several promising advantages, including:\n\n" \
+            "- **Enhanced robustness and generalization:** By removing the heuristic biases introduced by tokenization, these models could achieve better adaptability.\n" \
+            "- **Cross-modality scalability:** Since all data can be represented as bytes, these models naturally extend to multiple modalities.\n\n"
+        )
 
-        # Table header
-        f.write("| Title | Authors | Year | Link | Summary |\n")
-        f.write("|-------|---------|------|------|---------|\n")
+        f.write("This repository serves as an ongoing collection of papers and resources focused on byte-based LLMs.\n\n")
+
+        # papers
+        f.write("## Papers\n\n")
+        f.write("""The meaning of most fields are clear by their names. "Date" is the time that the work is released/made public (e.g., the timestamp of its first arXiv version). "Summary" is a one-sentence summary of the paper.\n\n""")
+
+        # table header
+        f.write("| Title | Date | Conference | Code | Summary |\n")
+        f.write("|-------|:----:|:----------:|:----:|---------|\n")
 
-        # Populate the table
+        # populate the table
         for paper in papers:
-            f.write(f"| [{paper['title']}]({paper['link']}) | {paper['authors']} | {paper['year']} | [Link]({paper['link']}) | {paper['summary']} |\n")
+            if 'code' not in paper or paper['code'] in ['N/A', 'NA', 'na', 'n/a', '-']:
+                code = '-'
+            else:
+                code = f"[Code]({paper['code']})"
+
+            if 'summary' not in paper or paper['summary'] in ['N/A', 'NA', 'na', 'n/a', '-']:
+                summary = '-'
+            else:
+                summary = "<details>" + paper['summary'] + "</details>"
+
+            f.write(f"| [{paper['title']}]({paper['link']}) | {paper['date']} | {paper['conference']} | {code} | {summary} |\n")
+
+        # contributing
+        f.write("\n\n## Contributing\n\n")
+        f.write("Contributions are always welcome. There are two ways to add a new paper:\n")
+        f.write("1. The easiest way is to open an issue, where I have a template for you to fill out.\n")
+        f.write("2. If you would like to be listed as a contributor, you add the paper to `papers.json` and make a pull request. Please do NOT directly edit `README.md`.\n")
 
 if __name__ == "__main__":
     generate_markdown('papers.json', 'README.md')